JP4001653B2 - Optical inspection of samples using multichannel response from the sample - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to automatic optical inspection of samples (eg, semiconductor wafers). In particular, in order to determine the presence of defects on or within the sample, a single search is performed during the inspection and therefore at least two independent inspection responses (eg bright field and dark field). Reflection from the sample at the same time. The inspection response can be collectively processed to determine the presence of defects.
[0002]
[Prior art]
Conventionally, there are three techniques for optically inspecting a wafer. In general, there are three methods using a bright field illumination method, a dark field illumination method, and a spatial filter.
[0003]
A technique that has already proved effective in inspecting pattern defects on wafers is the bright field illumination method using broadband illumination. This method can minimize the variation of contrast and coherent noise peculiar to a narrow band. The most successful example is the KLA model 2130 sold by KLA Instruments. The device can perform either a die-to-die comparison mode (die-to-die comparison) or a repeated cell-to-cell comparison mode (cell-to-cell comparison). However, the bright field illumination method may not provide sufficient sensitivity for small dust particles.
[0004]
[Problems to be solved by the invention]
In the image processing in the bright field observation, there is an event that a small dust particle diffuses a light beam to be concentrated on the aperture for condensing and reduces the energy of reflected light. It becomes the principle to detect. According to this principle, the energy loss generated is small when the dust is small compared to the point spread (point spread function) function value specific to each lens or the size of the pixel for digitization. The presence of garbage does not affect the observation results. Therefore, it is difficult to detect dust in bright field observation. If the generated energy loss is small, the location dependence of the reflectance of the surrounding bright region acts like a noise and hides the energy loss information due to dust. Naturally, if the location dependence of the peripheral reflectance acts like a noise, this is not true but it can be called a nuisance defect. In bright field observation, garbage is often hidden by false defects.
[0005]
In addition, even if there is a low-reflectance region around small dust particles, it is often seen on the wafer, and the reticle, mask and liquid crystal panel have essentially low reflectivity around the dust. Even if the reflected energy is small, the detection becomes even more difficult if it is further reduced by dust. Wafers, reticles, masks, liquid crystal panels, and other samples and other samples, therefore, employ dark field illumination methods in current commercial inspection instrumentation to detect small dust particles. In dark field illumination, a flat mirror will return very little signal to the detector, so the diffuse image is dark and hence the term dark field is used. On the other hand, any foreign matter protruding from the surface can return a large amount of light energy from the surroundings to the detector. It is understood above that an image usually looks dark except for dust particles protruding in the dark field or areas of circuit patterns having irregularities.
[0006]
The dust particle detector used in the dark field was designed and manufactured based on the simple hypothesis that dust particles scatter more light than circuit patterns. While this device works well with samples that do not have a pattern called a blank, when the circuit pattern is in the background, it can only detect large dust particles that protrude high above the pattern. As a result, the detection sensitivity naturally remains low, and is not satisfactory at the state-of-the-art advanced VLSI integrated circuit manufacturing site.
[0007]
In connection with dark fields, there are measuring instruments that specialize in a few unique aspects. A measuring machine of Company A is based on the principle of distinguishing from a circuit pattern by paying attention to the polarization characteristics of reflected light scattered by dust particles. That is, when incident light collides with dust particles, the reflected light exhibits a polarization state. However, if the pattern size becomes smaller than the wavelength of the incident light, the pattern size also imparts a polarization state to the reflected light in the same way as dust. This only increases false defects. Therefore, this method can be used only when the pattern size is large.
[0008]
Others, such as Company A and Company B, have introduced a mechanism that adjusts the angle of incident light, such that components of 0, 45, and 90 degrees are minimized by scattered light from the pattern. This is generally an improved dark field measuring instrument, but the scattered light from the corners of the pattern is still strong and gives a large noise component. In addition, in order to weaken this “pattern corner effect” and prevent false defects, it is necessary to deliberately reduce the detection sensitivity of a region having a high-density circuit pattern.
[0009]
Another method currently used to enhance the detection sensitivity of dust particles is to use a spatial filter. In plane wave illumination, the light intensity distribution at the back focal plane of the lens can be expressed by Fourier transform of the object placed there. In the case of a repetitive pattern, this Fourier transform becomes an array of light spots. If a filter is installed on the back focal plane of the lens to block the arrangement of the light spots, an appropriate spatial filter effect is given to the repetitive pattern and the repetitive component disappears. Therefore, only non-repetitive components, that is, only dust and defect images pass through. Company C, Company D, and Company E therefore primarily exercise spatial filter technology.
[0010]
On the other hand, the measuring apparatus based on the spatial filter technique has a big problem that it can inspect only a blank or an area having a repetitive pattern as described above. That is a fundamental limitation of this technology.
[0011]
The dark field spatial filter technology employed by a model of Company A uses a die-to-die image comparison and subtraction method for wafer inspection. With this technique, non-repetitive pattern areas on the wafer can be inspected by die-to-die comparison. However, even with the die-to-die comparison method, in order to obtain high sensitivity, it is still necessary to use a spatial filter in the repeated pattern region. In a high density memory cell, the signal from the normal internal circuit pattern in such a case is much stronger than the peripheral circuit and exceeds the appropriate dynamic response limit of the detection sensor. This leads to a saturation phenomenon in the sensor, and as a result small dust particles are not detected. Otherwise, the peripheral circuit will result in the signal from small dust particles being too weak. The problem is that in all cases small dust particles are not detected.
[0012]
In addition, Company A's dark field type spatial filter type die-to-die defect inspection apparatus has two major drawbacks. First, this device detects only dust particles and cannot detect any pattern defects. Secondly, in the case where an image passing through a spatial filter is normal, the screen is dark without a circuit pattern, so that it is impossible to accurately align the position for die-to-die. Of course, in order to exercise the comparison and subtraction algorithm, accurate alignment is essential. Company A tried to solve this problem by introducing an expensive high-precision stage mechanism. However, the introduction of such an expensive high-precision stage mechanism is not a satisfactory solution. This point can be attributed to the fact that there is no variation in the alignment of the pattern on the wafer due to a mechanical error during the stepper projection operation, and the synergistic effect of the residual error of the stage. Therefore, the results are not sufficient, and with this method of Company A, the size cannot fall below about 0.5 microns in a situation where small dust particles are not detected. Such a limit is caused by the fact that accurate alignment for die-to-die is not possible during inspection.
[0013]
There are several patent applications other than Company A, which are US Pat. No. 5,276,498, US Pat. No. 4,806,774 and US Pat. No. 5,177,559. However, with the exception of these listed prior applications, there is no precedent that has shown interest in the combination of bright field and dark field techniques because of a lack of understanding of the advantages of the techniques.
[0014]
Among the inspection apparatuses that enable observation of both bright and dark fields, all models on the market use a single light source that is common to both bright and dark fields. In addition, there is no technical method for determining defects by integrating inspection response signals from both bright and dark fields. Microscopes with both bright-field and dark-field illumination seen on the market have only a single light source that supplies both illuminations simultaneously, so it is not possible to separate the bright-field and dark-field observation signals from each other. Is possible. The only commercially available product from the Zuis microscope is that at least the bright field and dark field light sources can be separated at the same time, however, there is only a single detector and the method of separating the bright field and dark field illumination. There is no idea. There is only one dome light that can illuminate the whole sky.
[0015]
It is quite beneficial to have two separate illumination devices, bright field and dark field. Both advantages are directly linked to improved inspection equipment. Accordingly, it is an object of the present invention to provide an inspection method and apparatus discussed below. The present invention separately detects and correlates information from bright and dark fields, but this produces good results which are completely unexpected from the known examples.
[0016]
[Means for Solving the Problems]
The present invention relates to a defect inspection method characterized by referring to at least one known inspection response of a second pattern of the same design in order to search for a defect present at an inspection point of a first pattern on a sample, and Details of the device are described. In testing, at least one search is performed and accordingly at least two test responses are generated. It is important to use observation points that are equivalent to each other on the first and second patterns on the sample. The two responses (the response signals from the dark field and the bright field are representative) are detected separately by a photoelectric method, compared separately, and individually, a difference signal (between the first and second patterns). ).
[0017]
That is, the first and second responses in the first pattern are detected, and the result is compared between the two responses from the same inspection point in the corresponding second pattern, respectively, and the result As a result, a first difference signal and a second difference signal are formed. The differential signals produced individually are processed in order to determine a list of first pattern defects in a centralized manner. More specifically, the first and second differential signals can be collectively processed to determine a unified first pattern defect list.
[0018]
Data processing is then performed on the first pattern defect list. Then, false defects found on the known and harmless sample surface are extracted and eliminated. On the other hand, such known harmless false defects are provided to the user for reference. In addition, various inspection searches are added to increase the inspection response so that more than one optical response is obtained from the sample for processing. This further improves the inspection accuracy. In addition, in the case of a transmissive sample, a photoelectric detector can be installed behind it to collect the inspection response due to the transmitted light and improve the accuracy of the pattern defect list. It is also possible to search for defects buried in the surface.
[0019]
DETAILED DESCRIPTION OF THE INVENTION
According to historical facts, in the majority of defect inspection devices, the illumination means uses only either bright field or dark field, and there are no examples of using both illumination means. It can be said that. In a typical prior example seen from the historical background of such a defect inspection apparatus, only one illumination system of bright field or dark field is used at the time of observation as shown in FIG. In the prior art FIG. 1, the wafer 14 is illuminated with an illumination system having a suitable bright field 10 or dark field light source 12.
[0020]
Here, PLLAD (Phase Locked Loop Analog to Digital) is defined. This is a known method of A / D converting analog data into digital data by the Phase Locked Loop method, and this is hereinafter referred to as PLLAD. TDI (time delay integration) indicates a method of integrating signals by delaying signals. During operation, an image of the wafer 14 is captured by a sensor 16 that combines TDI and PLLAD as shown in FIG. 1, and a signal for carrying the image is input to an input buffer 18 (for example, RAM). Data is transferred to the defect detector 22 via the input buffer 18.
[0021]
The inspection data obtained from the die of the sample wafer by the defect detector 22 is compared with similar data of another similar die. Alternatively, the comparison partner may be from another reference wafer as long as it is the same kind of data from another similar die. This comparison operation is controlled by the delay circuit 20. Depending on the delay circuit 20 providing the correct delay timing, a die-to-die (die-to-die) or cell-to-cell (cell-to-cell) comparison operation is possible. Data output from the defect detector 22 is input to the post-processing processor 24. Here, an operation for determining the size and position of the defect is performed. Thus, it is possible to output a defect list with a threshold (for example, both KLA product models 2111, 2131 are designed to operate on the above principle using bright field light). .
[0022]
Consider changing the apparatus of FIG. 1 so that bright field and dark field observation results can be obtained independently and separately. This is a completely new approach and has never been done as a prior art, so such a system itself or its information cannot be obtained from others. The simplest and most obvious way is to perform such tests in a serial manner and in a manner that does not interfere with each other in order to implement such functions. In inspection 1, the light source 10 is used to provide a bright field. In another inspection, the light source 12 is switched to a dark field.
[0023]
According to a known technique, when the first inspection is performed with bright-field illumination, and in the next inspection, the wafer 14 is inspected with a dark-field light source 12 in a dark field, the sensor 16 detects a dark-field image of the wafer 14. The data passes through the buffer 18 and the delay circuit 20 and reaches the defect detector 22 and the post-processor 24. As in the case of the bright field image, it is possible to output the defect list 28 with a threshold value for the dark field image in this way. At this time, the post-processing processor 24 outputs independently the threshold value of the defect found in the dark field. That is, the conclusion here is that it is possible to output a defect list with a threshold value even in the dark field.
[0024]
Consider an inspection point on the wafer 14, that is, an image generation point. If this image generation point coincides with a defect generation point in the bright field defect list 26, the data value naturally exceeds the threshold value for a bright field defect. At this time, it can be said that the previous image generation point on the wafer 14 contains a defect. When the same operation is performed on the dark field separately from the above, if a certain image generation point on the wafer 14 coincides with a defect generation point on the dark field defect list, the data value is naturally a dark field defect. The threshold is exceeded. At this time, the previous image generation point on the wafer 14 contains a defect. Therefore, the image generation point on the wafer 14 may have only one of the bright-field type defect and the dark-field type defect, or it is possible to have both types of defects together. Both of the latter can happen. Thus, the post-processor 24 can output two independent and uncorrelated defect lists. One is based on the image from the bright field light source 10 and the other is based on the image from the dark field light source 12.
[0025]
FIGS. 2A and 2B are both examples that are known and show processing for determining whether or not the inspection data is defective. That is, the decision boundaries (34 and 40) in the figure need to be established in the bright field and dark field, respectively. Below the boundary line, as shown by reference numerals 32 and 38, the region which does not become a defect in the wafer 14 is shown. On the other hand, the area above the boundary line indicates a defective area as indicated by reference numerals 30 and 36. In developing the discussion of the present invention below, it should be pointed out that there is a problem in handling the known boundary as a line segment even though the boundary line indicating whether or not it is a defect cannot actually be represented by a line segment. . Therefore, FIG. 3 is referred to, and this figure forms a graph with the data step difference number in the bright field on the horizontal axis and the data step difference number in the dark field on the vertical axis. The data step difference number referred to here means that the signal intensity is first expressed in 256 steps in order from the lowest. Next, as described above, a signal strength comparison operation is performed, and numerical values indicate what kind of difference exists in the signal strength level of the data. Bright field and dark field threshold lines are drawn along each axis.
[0026]
Although there is no recording performed, it is possible to use both a bright field and a dark field with a combination of known techniques, and using this improves the accuracy in determining defects. Only when the data is plotted in a region where the data exceeds both the bright and dark visual field thresholds (the region indicated by reference numeral 38 in FIG. 3), it is possible to determine and identify this as a defect. I can do it. As an example of a device that can use both bright and dark fields simultaneously, there is illumination from the whole sky, that is, a dome type as shown in FIG. References introducing this are Yasuhiko Hara, Satoru Fushimi, Yoshimasa Ooshima and Hitooshi Kubota, "Automating Inspection of Aluminum Circuit Pattern of LSI Wafers", Electronics and Communications in Japan, Part 2, Vol. 70, No. 3, 1987. is there.
[0027]
Consider a dome-shaped lighting system with reference to FIG. Here, it is assumed that both illuminations are simultaneously performed on the wafer 14 using reference numerals 10 and 12 in FIG. Reference numeral 16 denotes a sensor which shares a single sensor with both bright and dark fields. It is not limited to the sensors to be shared between the bright and dark fields, and both visual fields, and the same applies to the process equipment groups 18 to 24 (processor channels). As a result, a single output is realized, and the result is shown in FIG. It should be noted here that FIG. 2C does not mean that the separated inspection is performed twice in the bright field and the dark field. Here again, the problem is that the threshold value is determined linearly and is different from the reality.
[0028]
FIG. 4 is publicly known. If the dome-shaped illumination is a first modification, this is a second modification of the defect search apparatus. Here, the defect search can be executed simultaneously in the bright field and the dark field. In FIG. 4, there are two sets of processor channels ranging from 18 to 24, one of which is dedicated to bright field and two of which is dedicated to dark field. Now consider the light source in this case. There may be a single light source or a pair of light sources. The paired light sources refer to the case where one light source is set for each field of view, and a total of two light sources are set. Two light sources can be used simultaneously as a pair of light sources in sequence and as a dome type illumination. In either case, bright field and dark field illumination light is applied to the wafer 14.
[0029]
The difference between FIG. 4 and FIG. 1 is whether to have only one or two processor channels. Even with this, even if there is a difference between sequential or simultaneous, it is possible to perform a bright and dark field inspection. The number of processor channels held is not an essential difference. Therefore, the bright field defect list 26 and the dark field defect list 28 can be obtained independently of each other in the simultaneous inspection or the separate inspections. If the channels of each processor are operated in synchronization with each other, the speed of the slowest channel can be met. Therefore, the sweep for capturing an image on the sample only needs to be performed once at the synchronous speed. As a result, defect list data for both bright and dark fields and both visual fields can be obtained at once.
[0030]
Therefore, the system shown in FIG. 4 is more effective than the one shown in FIG. 1 (because only one sweep is required). In the following, the defect list data for both light and dark and both visual fields will be referred to as an alias map. Here, the data processing can be performed in parallel with the sample alignment operation, which is very convenient. However, both the known system of FIG. 1 and the known system of FIG. 4 adopt the simple line segment method described above in addition to determining the defect threshold values (lists 26 and 28) independently. Because of this, it is different from reality. This is a problem. That is, all the discussions regarding the problems of the simple line segment method apply to each of FIGS. 2 (a), (b), and FIG.
[0031]
Therefore, the contents of the present invention will be described in detail below with reference to FIG. This is a block diagram of the present invention. Sensors 16 and 16 'are provided, except that the image of the wafer 14 is dark and dark, and that each field of view data is captured separately, the left side of the diagram of FIG. Is identical to the part. Here, naturally, the sensors 16 and 16 'detect the bright-field and dark-field signals separately, and then input these signals separately to the buffers 18 and 18', and further pass the signals through separate delay circuits. Send it first. The resemblance between the present invention and the known example of FIG.
[0032]
A single defect detector shown at 22 in more detail in FIG. 6 receives both the bright and dark signals and the two visual field signals which are separately input to the buffers 18 and 18 'and sent through the separate delay circuits. The defect detector 22 performs an operation necessary to know the position of the defect on the wafer 14. The defect detector 22 outputs a highly consistent defect list that combines light and dark and both fields of view comprehensively. The post-processing processor 42 receives this data and determines whether the captured defect is a pattern defect 44 or a dust particle 46.
[0033]
In FIG. 6, the post-processor 42 can use a VME (Virtual Machine Environment) bus process board based on the Motorola 68040 CPU having high performance and versatility. Alternatively, an example known as a KLA Model 2131 high performance post-processor board may be used. As is well known, the surface pattern of a semiconductor wafer may change depending on the location when viewed in contrast. Sometimes it has a grain-like pattern called grain. Moreover, grains may be formed in a lump. In addition, the manufacturing process frequently changes. Irregularly shaped patterns (marks) due to chemical causes also adhere. Since these are not problems of yield and reliability, it is not necessary to recognize them as defects here. Each of the patterns associated with the surface variations, when viewed during inspection, causes the images in the bright field and dark field to emit a typical range of signal values. In addition, it is necessary to consider noise. When the signal value is small, noise adversely affects the signal, particularly light and dark, and the step difference value in each field of view. Thus, for a typical example, the relationship between light and dark, the number of step differences in each field of view, and system noise was plotted. This is FIG.
[0034]
In FIG. 7, the system noise 54, the surface contrast fluctuation 56, and the grain 58 are all close to the origin, and their values are small in both bright field and dark field. On the other hand, the process variation is 75% in the bright field and about 50% of the midpoint in the dark field. In addition, when grains are formed in a lump, the brightness and darkness and both fields of view are large. Ideally, the best system is such that these predictable and harmless fluctuations can be eliminated without recognizing them as defects, and only the signal shown at 48 in FIG. 7 is recognized as a defect. Say something.
[0035]
FIG. 6 is a partial block diagram of the circuit shown in FIG. Here, the defect detector 22 is shown in detail. The defect detector 22 is briefly represented in FIG. 6, but when receiving the input signal, the signal passes through the input buffers 18 and 18 'and the delay circuits 20 and 20'. The defect detector also includes filter groups 90 and 92, and 90 'and 92'. Filter groups 90 and 92 and 90 'and 92' process image data. At this time, 3 × 3 or 5 × 5 pixels are digitally filtered. There is also an example in which KLAInstruments product Model 2131 is applied. The image signal is preprocessed by the filter groups 90 and 92 and 90 'and 92', and then input to the subtracters 94 and 94 '. Here, a comparison operation (subtraction) for each image pixel is performed for comparison with the corresponding delay data. In a typical example, in the case of die-to-die comparison (die-to-die), the delay time difference corresponds to the width of one die. In the case of cell-to-cell comparison, the delay time difference corresponds to the width of one cell, and the same delay amount is applied for each field of view. Accordingly, the outputs from the subtracters 94 and 94 'are, as a matter of course, defect information for the wafer 14 in both the light and dark fields and in both fields of view. The information can then be alternately supplied to the two-dimensional histogram circuit 96 and the post-processor 42. On the other hand, the coordinate value of the data in FIG. 7 is given to the signal component directly input to the post-processor 42. Then, the two-dimensional histogram circuit 96 separately forms a two-dimensional histogram diagram of defects in both the bright and dark visual fields. The two-dimensional histogram signal obtained by such processing is supplied to a determination algorithm 98 for determining whether or not it is a defect. And known and frequent non-hazardous wafer surfaces and other signal due to other variations (such as system noise, grain, contrast variation, process variation, grain clumps, etc., as described above) Is a non-defect, and hence is referred to as a “false defect”.
[0036]
The predictable and harmless false defects obtained by the prior art methods of the prior art should not be confused with true defects. FIG. 8 shows what should be done for this purpose. That is, in FIG. 8, as described above, the threshold value determined by the straight line indicated by 34 or 40 does not apply to an actual defect, but this is because there is no other method in the known example. I have to adopt it. As can be clearly understood from the above, the region 38 is a proposal as a defect region in which the data of both the bright and dark fields of view are comprehensively integrated, but is considerably small and inappropriate. This is not suitable for actual use. Refer to FIG. 7 again. In the present invention, the defect detector 22 can be programmed in various ways because it independently and simultaneously collects and processes the data of light and dark and both visual fields. Therefore, a complex nonlinear threshold function can be applied to the difference in the number of steps in both the bright and dark data and the visual field to eliminate false defects, and only the black part excluding the white area in FIG. Can be separated as a defect signal.
[0037]
In this way, false defects are safely eliminated. In other words, in the present invention, it is possible to handle both bright and dark signals and signals of both visual fields at all the step difference numbers from 0 to 255, and the false defects 50 and 52 shown by the white dots in FIG. 54, 58 and 56, etc. can all be avoided.
[0038]
Next, one physical optical embodiment of the present invention is shown in FIG. The wafer 14 is directly illuminated by a dark field light source 12 (eg, a laser) and a bright field light source 10 (eg, a lamp based on the mercury arc principle). At this time, the light beam passes through the two lenses 60 and 62 and further through the beam splitter 66. The light beam carrying the image signal of the wafer 14 depending on the bright field and dark field integrated in this way passes through the condenser (compression) lens 60 and passes upward, and passes through the beam splitter 64 and the beam splitter 66. The bright field light passes through the beam splitter 66, travels further upward, passes through the condenser lens 72, and enters the bright field sensor 16. On the other hand, dark field light is reflected by a film (dichroic film) coated with a dichroic film on the beam splitter 66. This is because the wavelengths of the light sources in the dark fields of view are different from each other so that they can be designed to do so. The dark field light passes through the spatial filter 68 and enters the relay lens 70, and then reaches the dark field light sensor 16 '.
[0039]
As used herein, dark field illumination consists of a combination of a laser and a spatial filter 68. Therefore, in this system, a Fourier transform plane of the image of the wafer 14 can be formed. The purpose of using the spatial filter 68 is to selectively remove normal patterns having no defect. It is pointed out that this improves the accuracy of defect search. As described above, the present invention uses a mercury arc light source that passes through the beam splitter 64 in the bright field and a laser light source that allows spatial filtering in the dark field and two separate light sources.
[0040]
Here, since the laser light source has high brightness and high output, even if the information of the bright field and the dark field are handled separately, the light loss is small, which is only a few percent. Considering that a narrow-band laser light source is sufficient for dark field illumination, a 633 nanometer wavelength helium-neon laser can be used. Alternatively, a laser diode with a wavelength of 630 to 830 nanometers can be used. Then, only the dark field light can be reflected well by the dichroic coating material (dichroic film) on the beam splitter 66 described above. Alternatively, dark field light can be easily separated by using an Oriel model 52720 with a laser interference filter. When combined with the latter, a narrow band spectral filter, the Oriel model number 56460 can be used for bright-field light with a mercury filter. Alternatively, Oriel has a custom design of narrowband notch filters for laser applications that can be used. As a result, spatial filtering is only used for dark field optical paths. In this way, the bright field optical path is not affected and the quality of the image is kept high. When using a narrowband light source (above described an example using a narrowband laser light source for dark field illumination), a spatial filter operation is required. The narrow band characteristic peculiar to laser light is generally easy to separate bright-field and dark-field signals from each other with a filter and a beam splitter. The spatial filter 68 inserts and exposes a piece of photographic negative film as shown in FIG. The spatial filter can be manufactured by removing the film after exposure and developing it. Of course, after development, the film itself becomes a filter, so it is placed in the correct place shown as 68 in the figure. In addition, the spatial filter 68 can be realized by using a field effect filter SLM (Spatial Light Modulator), that is, a liquid crystal LCD (Liquid Crystal Display) manufactured by Hughes.
[0041]
It has been stated that the preferred embodiment has the various options when only the dark field signal is well separated from the entire image signal beam. However, it is pointed out in the present invention that the reflection effect based on the film (dichroic film) coated with the dichroic film on the beam splitter 66 is combined with the spatial filter 68 to be most effective overall. To do. The reason for this is well understood when considering the controllability and wide dynamic range. Since the dark field signal is processed simultaneously with the bright field signal, it is particularly important to consider this controllability and the wide dynamic range.
[0042]
The method of using the reflection effect due to the dichroic film on the beam splitter 66 has been presented. However, with the current development of optical technology, it is quite possible that there are still unknown methods with advanced technology.
[0043]
FIG. 10 shows the principle of the second embodiment of the present invention. In this embodiment, there is a single laser 76 that provides both bright field and dark field illumination to the wafer 14. Here, the beam splitter 80 reflects light downward, and the light beam enters the wafer 14 via the condenser lens 78. In the present embodiment, the image processing is performed at the same time in each of the bright and dark fields. The dark field detector 74 is installed at a small elevation with respect to the wafer 14, while the bright field sensor 82 is installed directly above the wafer 14. At this time, the light beam passes through the condenser lens 78 and the beam splitter 80. In order to optimize the efficiency of defect search in this embodiment, the output of the bright field detector 82 and the output of the dark field detector 74 are preferably processed simultaneously for the target defect. Using a bright field image set in a wide band and a dark field image formed using a spatial filter for die-to-die comparison (die-to-die) is a legitimate method and overcomes all the limitations of previous inspection equipment To do. Since there is a bright field image, it can be used to conveniently align two dies for comparison with high accuracy. If the sensors in the dark and the visual field are aligned (pre-aligned) in advance, the same region can be observed and inspected accurately. Then, it is only necessary to measure the misalignment between the two dies to be compared with the bright field channel, and the alignment is automatically performed with the dark field channel.
[0044]
Adjustment and calibration are performed at the time of manufacture, and the misalignment (offset) of the alignment is fixed between the dark and light fields, as is obvious from the above. The amount of offset is stationary and device specific and it is also a known value. Thus, a set of high speed electronics systems for measurement for offset need only be provided for the bright field, and is therefore unnecessary in the dark field channel. If the alignment information obtained from the bright field image is used, even in the dark field channel, a fairly accurate die-to-die alignment is performed for the above-described reason, so that the alignment residual error is sufficiently small and small dust particles on the sample. There will be no hindrance to the search.
[0045]
Based on the above, in the data processing of the information obtained in the dark field using the spatial filter, at present, most repetitive patterns on the die and areas with many straight lines are excluded by the filter. It is desirable to adopt a method. As a result, the dynamic range (operating range) becomes uniform, so that small dust particles and the like can be captured in one search regardless of whether the pattern is dense or sparse. In addition, more information about the defect can be obtained by the technique of simultaneously processing the dark field image and the bright field image. For example, a bright-field image makes it possible to simultaneously capture both pattern defects and defects such as small dust particles, but in contrast, dark field images can detect only defects such as small dust particles. It is possible. Therefore, if the difference between these two results is taken, the answer is only a pattern defect. The ability to automatically separate defects such as small dust particles described above from pattern defects in real time can be said to be a very characteristic ability in the technology of the present invention. The value of this for wafer inspection is incredible. In this special application, with respect to defects such as small dust particles, the dark field image has a much higher detection sensitivity than the bright field image. Therefore, matching with the bright field image can be achieved by reducing the detection sensitivity of the dark field image. As a result, it can be said that the defect obtained from both the bright and dark channels is a defect such as a small dust particle, and the pattern defect can be obtained only from the bright field image.
[0046]
Another example will be described. It is an example of inspection at a metal wiring layer in a semiconductor wafer. Combining the results obtained from dark and both field images, it is possible to successfully separate the false defects created by the grain of the metal wiring (the grain pattern on the surface) from the true defects. This need is great and therefore important. Both bright and dark images, images of both fields of view and delayed images corresponding to them are received and stored separately. Thereafter, as shown in FIG. 5, the die-to-die alignment is performed, the stored data is taken out and inputted to the defect detector 22. In order to perform the detection operation in this way, a gigabyte class dynamic memory (DRAM) is required. The dynamic memory stores the detected data, and the data is read from the memory and used as necessary while considering the timing as described in FIG. Although the system operates in this way, the system shown in FIG. With today's technology, it is preferable to inspect images of bright and dark fields and both fields of view in real time in wafer inspection, which is not only preferable, but also cheaper. Either method can be used, but it is of course important to obtain images of bright and dark fields and both visual fields at the same observation point on the wafer 14 and input them to separate detectors. . It is very important to recognize the fact that the inspection is performed on exactly the same observation point on the wafer 14, and based on this recognition, the relationship between the bright and dark images and the two visual field images becomes clear. (For example, if the dark field image is very bright, but the bright field image is quite dark, you understand what this means.)
If the above-mentioned two image signals in the bright and dark fields and both visual fields are simply added first and then input to one detector, the same result as in the present invention will not be obtained. The reason will be described below. If the latter is ignored ignoring the excellent results of the present invention, the difference between the two image signals in both the light and dark fields and both fields of view is canceled out and only the important information and ability to detect the defect is lost. It ends in. This is where the true value of this invention exists. Brightness and darkness, each field image can set illumination light to completely different light sources. In addition, the incident angles of the light rays from the light source are supplied while being changed considerably from one another. Since the obtained images are input to separate detectors, the actual situation on the wafer 14 can be fully depicted. In order to carry out this operation, the two detection sensors need to be linked with each other so that the optical axes are correctly aligned. However, linking such detection sensors to each other and aligning the optical axis leads to the opposite result in terms of reducing the cost of the defect search apparatus and increasing the complexity. On the other hand, there is also a merit of the present invention described above. Since there are no known examples, the advantages of the present invention are difficult to understand.
[0047]
Further, in the discussion so far, it has been assumed that light having a single frequency (single wavelength) is used for a light source for a bright field or a dark field for detecting a defect. However, there is no reason for this technology to remain. It is correct to think of further expansion, and doing so can increase the channel of information further. That is, a multi-channel can be constructed using light sources having a plurality of frequencies in the bright field or dark field. The key to this expansion is the two-channel technology described above, which is of course fundamental. If the point is repeated once more, it is first mentioned that the same observation point of the wafer 14 is used. The next step is to perform photoelectric data processing with a separate detector for each field of view. Then, the detection results are arranged as shown in FIG. 7 (here, the case of two channels was discussed, but the same applies to multiple channels). In the case of multi-channels of two or more channels, FIG. 7 is naturally not multi-dimensional but multi-dimensional. It is almost impossible to properly express more than three dimensions on paper. However, in the multidimensional case, it is known that a numerical method based on a computer is useful, and it can be used.
[0048]
FIG. 11 shows a system obtained by developing FIG. In this example, a multi-channel bright field and a multi-channel dark field are used. Specifically, two channels are used for a total of four channels. Accordingly, FIG. 11 is based on FIG. 9, but repeated description of FIG. 9 is avoided here. However, all the elements shown in FIG. 9 function as they are in FIG. In the second dark field channel, the laser 12 ′ oscillates at a different frequency than the laser 12, but is introduced in addition to illuminate the same location on the sample wafer 14. The detailed description of observing the sample wafer 14 will be continued. In FIG. 11, a second bright field illumination is further added, specifically the light source 10 ′. This is naturally set so as to emit light having a frequency different from that of the light source 10. The lens 62 ′ and beam splitter 64 ′ are required to direct bright field illumination to the same location on the sample wafer 14. Even in the reflection mode, there is no significant difference in operation content from that of FIG. A beam splitter 66 'using a film (dichroic film) coated with a dichroic film is further added so as to reflect the light from the second laser 12'. This reflected light is incident on the detector 16 ″ via the spatial filter 68 ′ lens 70 ′. In addition, the beam splitter 73 having a film (dichroic film) coated with a dichroic film is a bright field illumination light source 10. Alternatively, any one of 10 'is selected and set to reflect so that the reflected light is incident on the detector 16'''. At this time, the bright-field illumination light beam that passes through the beam splitter 73 is different in frequency from the reflected light. The bright field illumination light beam that passes through the beam splitter 73 can be set to be separated when passing through the beam splitters 66, 66 'and 73'.
[0049]
The embodiment of FIG. 11 is merely a modification of the embodiment shown in FIG. The present invention can provide a wider range of examples. When detecting special defects separately from other defects, it is necessary to combine various elements. The embodiment in which various elements are combined in this way is different from that described here, but the concept may be considered the same. When the information obtained from the multiple channels is used to determine that the obtained signal is “defective”, the present invention is characterized in that the signal information from both the dark and the visual fields is used simultaneously. In all the known examples, the information channel is limited to a single one. On the other hand, in the case of multiple channels, as described above, a plurality of wavelengths can be set in both bright and dark visual fields, and a corresponding light source can be used. Moreover, the technique used here for wafer defect search can be further expanded, for example to search for defects on transparent materials. In such an application, bright field light and dark field light depending on the transmitted light are incident on the detector, and on the other hand, there is also light that is reflected, so the electric signals of the bright field light and dark field light depending on the reflected light are integrated. It is also possible to search for various defects and further identify their positions. FIG. 12 shows an embodiment used when executing the above, but the actual state is shown in a simplified manner.
[0050]
FIG. 12 and FIG. 9 are almost the same, and the difference is that the embodiment of FIG. 12 is constructed by reproducing the latter optical signal detector portion below the sample 14 '. The light beam carrying the image information of the combined transmitted light in the bright and dark fields of view travels further downward from the bottom surface of the sample 14 'and reaches the beam splitter 66T through the compression lens 60T. At the beam splitter 66T, the bright-field image signal light beam travels further downward to reach the compression lens 72T, and accordingly is set to be projected onto the bright-field light sensor 16T dedicated to transmitted light. On the other hand, only the image of the transmitted light, particularly dark field light, can be set to be reflected by a film (dichroic film) coated with a dichroic film on the beam splitter 66T. This is because the wavelength of the illumination light source for each field of view is generally different. The remaining dark field light passes through the spatial filter 68T, passes through the relay lens 70T, and enters the dark field light sensor 16T '.
[0051]
In the above, the specific case of each bright-field and dark-field illumination source has been described for the inventive concept. That is, the light irradiating the sample uses an illumination light source in each of bright and dark fields, and the resulting signal and detection are performed independently. Generally speaking, the present invention can be configured to include all of the following elements a), b) and c). a) There is at least one image search mechanism (probe) that is deployed to have at least two independent optical response systems in searching for the same location of the same die of the sample to be examined. . If more than one probe is used, all probes should be deployed to search for the same location on the same die of the sample to be examined. b) In order to perform a search operation through each optical response system, as described above, an operation for comparing with image data at equivalent positions of other dies is performed. Is generated. c) Further, the same operation is performed for the other optical response systems, and a multiple operation is performed to create a first list of pattern defects based on the same principle. Brightness and darkness, I mentioned above on how to use each field of view, and discussed extending it to a generalized process. Now, post-processing can be performed on the obtained first list of pattern defects as follows. That is, this post-processing should be performed on known false defects on the surface. False defects are not “true” defects, as mentioned earlier. Therefore, it should be excluded from the final pattern defect list. This post-processing makes it possible.
[0052]
In the discussion made on the drawings cited above, one or more probes can be provided to generate one or more sets of optical responses. In FIG. 10, the probe is single. The laser 76 is used as an illumination source for generating both bright and dark field images of the sample. In addition to introducing the wafer 14, two independent detectors, one dedicated to the dark field detector 74, and the other dedicated to the bright field detector 82, the information signal system has two channels. I have to. In FIG. 9, there are two probes. The laser 12 serves as a common illumination light source for two different dark field systems, and irradiates the sample way 14. Separately, the lamp 10 is a light source of a bright field system. There are two independent detectors 16 and 16 ', which receive the reflected light from the sample in both bright and dark visual fields to form two channels. FIG. 11 is obtained by extending the system of FIG. Here, a second dark field light source and a bright field light source are respectively added. The result is 4 probes. For the detector, 1 was added to each of the bright and dark visual fields to make the information system a total of 4 channels.
[0053]
On the other hand, FIG. 12 is the same as FIG. 9, but the number of probes is two, and an illumination system is provided for each of the bright and dark visual fields. In addition, a transmitted light detector is provided for the bright field and a radiant light (reflected light) detector is provided for the dark field, for a total of 4 channels. That is, there are two channels of transmitted light and reflected light in a bright field, and two channels of transmitted light and reflected light in a dark field. In each of the above examples, there is no shift in frequency and phase of the optical signal. That is, there is no difference in frequency and phase between the light emitted from the probe and the light incident on the detector. The event that occurs is the separation of the optical signal beam.
[0054]
Fluorescence is one of the photoreactions, and is a phenomenon that is widely known for certain materials, and occurs when a material is irradiated with light of a certain wavelength band. When the material emits fluorescent light (which is secondary light), the frequency (wavelength) of the light is smaller (longer wavelength) than the primary light, ie, probe light. For some types of materials, it may be effective to observe a shift in frequency associated with the generation of fluorescence in the defect search. The frequency of fluorescence in light emission is unique to the material and is well studied. The applied dichroic film (dichroic film) used for the beam splitter in the present invention can select a material sensitive to the frequency of incident light, and the detector can also do so. A beam splitter and detector that are sensitive to the frequency of the incident light, as well as other effective techniques, can be introduced into the system of the present invention.
[0055]
Similarly, the illumination light that reflects when the optical path from the probe to another inspection area on the sample (there is a step on the silicon wafer surface or the refractive index of the light may be different in a remote area on the wafer) is different. Produces a phase shift (phase difference) with the light source. The phase difference information is useful for certain defects, so it is effective to introduce a phase difference detector in one channel. If an interferometer is used, phase difference detection is easy. The interferometer can also detect contrast differences on the sample. There are many types of interferometers. Mach-Zehnder, Mirau, and Jamin-Lebedeff, and others that use the light beam twist principle, are all usable in this invention. In addition, it is possible to observe in consideration of the gradient value of the phase change. For this, a differential method, ie, a differential or Nomarski type contrast microscope can be used. Depending on the state of the sample, there may be a change in the polarization state especially with the change in the phase of the observation light. This is also an information source for defect search.
[0056]
By way of example, transmitted light carries that information when the sample has a double refraction and it has a position dependence of the observation region. Similarly, the sample may have reflection characteristics sensitive to polarization or scattering characteristics. At these times, the reflected probe light has the characteristics. The above-mentioned fluctuation (shift) of the polarization state in the probe light can be detected by a conventional instrument, and easily provides a new information channel in the sample inspection process. Hey. Whether the irradiation light to the sample is applied from above or from below can be determined from design factors that also consider the incident angle. Focus-sharing (confocal) illumination can be used as probe light. At this time, the topography (surface shape) information on the sample can be observed, and a new information channel can be formed.
[0057]
A technique different from that described above when searching for defects using probe light will be described. There may be other techniques that have not yet been discovered. One of them is a technique for observing time-varying information using pulse-type irradiation light (selecting the pulse shape of the probe light, that is, creating a pulse pattern by the on-off operation). The above-described dynamic signal that fluctuates over time can be used for organizing / editing detection information or organizing / editing multiplex information, thereby simplifying the defect information detection process. It is possible to detect the topographic characteristics of the sample surface by observing the time shift (time delay) seen in such a dynamic signal. Some commercialized cameras have multiple sensors housed in the same housing. A camera having each color element of RGB (red, green, and blue) is an example, and three independent CCDs (charge coupled devices) are provided in one camera housing. With such a camera, the optical axes of three independent CCD sensors can be automatically aligned with a single adjustment. Each sensor is an individual sensor in such a camera and has an individual signal processing mechanism. In each embodiment of the present invention, it is absolutely necessary to align the plurality of probe mechanisms so as to aim at the same point of the sample. Of course, it is absolutely necessary to perform the same alignment in the detection mechanism, and the image size at the time of image capture must be the same.
[0058]
The various operation modes of the present invention have been described above. The contents are typical settings and technical examples and details based on a model example for continuing the inspection. When an expert carefully reads this specification with reference to the drawings, a modified example to which the present invention is applied will immediately come to mind. Therefore, the claims set forth herein should be considered to include all modifications, applications and variations thereof. In short, the claims should be interpreted appropriately and broadly based on the true spirit and vision of the technology based on this invention.
[0059]
【The invention's effect】
As described above, according to the present invention, the first and second responses in the first pattern are detected, and the result is compared between the two responses from the same inspection point in the corresponding second pattern, respectively. As a result, first and second differential signals of the response are formed. These individually generated differential signals are collectively subjected to data processing, and a unified first pattern defect list is generated. Data processing is then performed on the first pattern defect list. The false defects found on the known and harmless sample surface are then extracted and eliminated. In addition, various inspection searches can be added to increase the inspection response and obtain and process more than one optical response from the sample. This further improves the inspection accuracy. In addition, in the case of a transmissive sample, a photoelectric detector can be installed behind it to collect the inspection response due to the transmitted light and improve the accuracy of the pattern defect list. It is also possible to search for defects buried in the surface.
[Brief description of the drawings]
FIG. 1 is a block diagram of an optical inspection system according to a single channel response method in a prior inspection technique, which sequentially inspects a defect of a wafer in a bright field and a dark field.
FIG. 2 (a) is a result of an inspection using a bright field in the prior inspection technique, and a graph in which a threshold is determined and a signal having a signal intensity exceeding the threshold is classified as a defect. FIG. 2 (b) is a result of inspection using a bright field in the prior inspection technique, and a graph in which a threshold value is determined and a signal having a signal intensity exceeding that value is classified as a defect. FIG. 2 (c) is a graph showing the result of the inspection using the dome illumination system in the prior art, in which a threshold is determined and a signal having a signal intensity exceeding that value is classified as a defect.
FIG. 3 shows the result of wafer inspection in the prior inspection technique, in which the step difference in signal data in the bright field is plotted on the horizontal axis and the step difference in signal data in the dark field is plotted on the vertical axis. A graph in which a value is determined and a signal having a signal strength exceeding that value is classified as a defect.
FIG. 4 is a block diagram of an optical inspection system in a prior inspection technique, in which a wafer is inspected for defects in a bright field and a dark field by two independent signal processing systems (processor channels). .
FIG. 5 is a block diagram of an automatic optical inspection system according to the present invention, in which a bright field and a dark field are input to the same processor channel in order to integrate a bright field and a dark field, respectively, and a wafer is inspected for defects.
6 is a diagram showing in detail a defect detector in the automatic optical inspection system shown in FIG.
FIG. 7 is a graph in which the number of step differences in bright field is plotted on the horizontal axis, and the number of stage difference numbers in dark field is plotted on the vertical axis, showing the relationship between intrinsic defects, false defects, and noise.
FIG. 8 shows how each line segment showing its bright and dark field thresholds is far from the actual in the prior art by plotting a combination of FIG. 3 and FIG. Graph.
FIG. 9 is a schematic diagram showing, in a simplified manner, the principle of using separate illumination systems in the bright field and dark field according to the first embodiment of the present invention.
FIG. 10 is a schematic diagram showing the principle of using a single same illumination system in a bright field and a dark field according to a second embodiment of the present invention.
11 is a third embodiment of the present invention, partially similar to FIG. 9, with two illumination systems and two detector subsystems in the bright field, and also a dark field. FIG. 2 is a schematic diagram of a system having two illumination systems and two detector subsystems.
FIG. 12 is a schematic diagram of a system according to a fourth embodiment of the present invention, in which light is transmitted through a transparent sample, and separate illumination systems are used for bright field and dark field. .
[Explanation of symbols]
10 ... Bright field illuminator
12 ... Dark field illuminator
14 ... Sample wafer
16 ... sensor
18 ... Input buffer
20 ... Delay circuit
22 ... Defect detector
24. Post-processing processor
26 ... Bright field defect list
28 ... Dark field defect list
30: Possible value of signal
32: Possible value of signal
34 ... Line indicating threshold
36: Possible value of signal
38: Possible value of signal
40 ... Line indicating threshold
42. Post-processing processor
44 ... pattern defect
46 ... dust particle defect
48 ... defects
50 ... grains
52 ... False defects due to process variations
54 ... System noise
56 ... False defects due to surface contrast fluctuation
60 ... Lens
62 ... Lens
64: Beam splitter
66 ... Beam splitter
68. Spatial filter
70 ... Relay lens
72 ... condenser lens
73 ... Beam splitter
74. Dark field image detector
76 ... Laser
78 ... Condenser lens
80: Beam splitter
82 ... Bright field image detector
90 ... Filter
92 ... Filter
94: Subtractor
96 ... Histogram circuit
98 ... Defect determination algorithm

Claims (34)

A method for searching for defects existing in a first pattern on a sample with reference to a known dark, reflected image in both fields of a second pattern made with the same design on the same sample,
a) illuminating the same point of the first pattern with a bright-field light source and a dark-field light source, respectively;
b) detecting a dark-field reflected image obtained by reflection from the first pattern;
c) detecting a bright-field reflection image obtained by reflection from the first pattern;
d) comparing the dark field reflection image obtained in step b) with the dark field reflection image at the same point of the second pattern to generate a difference signal of both images;
e) comparing the bright field reflected image obtained in step c) with the bright field reflected image of the second pattern to form a difference signal between the two images;
f) a difference signal of step d) and in the reflection image on both the field of view of the dark-light obtained in e) and the data processing comprises a step of determining one of the defect list centrally first pattern
The step of data processing the difference signal of the reflected image in both the dark and bright visual fields is as follows:
1) generating a two-dimensional histogram based on the difference signals of the dark and bright visual fields;
2) identifying a region in the two-dimensional histogram associated with a process or surface variation of the sample; and
3) excluding such regions from the combination of data representing a two-dimensional histogram; and
4) determining a defect list for the pattern based on data from the remainder of the combination of data and not based on data from the excluded region of the two-dimensional histogram. And automatic defect inspection method. The automatic defect inspection method includes:
g) adding post-processing to the data of the first pattern defect list obtained in the step f), and finding a harmless false defect in the list and eliminating the false defect data. Item 2. An automatic defect inspection method according to Item 1.   2. The automatic defect inspection method according to claim 1, wherein the illumination light source in step a) is composed of an independent dark field illumination light source and bright field illumination light source.   4. The automatic defect inspection method according to claim 3, wherein the light of the light source for each of the dark and visual fields independently formed is configured to have a different frequency. The dark field light source is arranged to provide illumination of narrow band light and the step b)
5. The automatic defect inspection method according to claim 4, further comprising the step of: h) enhancing the defect search capability by the light beam of the dark field image passing through a spatial filter. The sample is optically transparent, with reference to the known dark, transmitted image in both fields of the second pattern;
The method
i) detecting a dark field transmission image of the first pattern;
j) detecting a bright-field transmission image of the first pattern;
k) comparing the dark field transmission image obtained in step i) with a dark field transmission image at the same point of the second pattern to form a differential signal of both images;
l) comparing the bright field transmission image obtained in step j) with the bright field transmission image at the same point of the second pattern to form a differential signal of both images,
The step f) includes the dark field and the bright field obtained in the steps d) and e), respectively, in addition to the difference signal relating to the transmitted images in the dark field and the bright field obtained in the steps k) and l). The automatic defect inspection method according to claim 1, further comprising: a step of performing data processing on a difference signal relating to both reflection images at a step to determine a defect list of the first pattern in a unified manner.   The automatic defect inspection method includes the step of m) performing post-processing on the defect list data of the first pattern obtained in step f), and searching for surface fake defects whose characteristics are known. The automatic defect inspection method according to claim 6.   7. The automatic defect inspection method according to claim 6, wherein said step a) provides independent dark field and bright field illumination light sources.   9. The automatic defect inspection method according to claim 8, wherein each of the darkness and the illumination of each visual field is composed of light sources having different frequencies. The dark field illumination source is arranged to emit narrowband light;
Said step b)
n) enhancing the defect search capability by passing the light beam of the dark field reflected image through a spatial filter;
10. The automatic defect inspection method according to claim 9, further comprising the step of: o) enhancing the defect search capability by passing the light beam of the dark field transmission image through a spatial filter. The second pattern on the sample is set with a known reflected image based on a dark field light source emitting a plurality of lights each having a different frequency as a reference image,
Said step a) illuminates the same observation point of the sample according to the setting of the dark field based on multi-channel light sources of different frequencies;
The step b) separately detects reflected images obtained by irradiating the same observation point of the first pattern in each of the multiple channels.
In step d), the multiple reflection image obtained in step b) is compared with a known reference image of the second pattern, and a difference signal between the two reflection images in the dark field is generated. The automatic defect inspection method according to claim 1. The second pattern on the sample is set with a known reflected image based on a bright-field light source that emits a plurality of lights each having a different frequency as a reference image,
Said step a) illuminates the same observation point of the sample according to the setting of bright field based on multi-channel light sources of different frequencies;
The step c) separately detects reflected images obtained by irradiating the same observation point of the first pattern in each of the multiple channels.
The step e) compares the multiple reflection image obtained in the step c) with the reference image of the known second pattern described above, and generates a difference signal of both reflection images in the bright field. The automatic defect inspection method according to claim 1, wherein: It is an automatic defect detection device that searches for defects existing in the first pattern on the sample with reference to the known dark and reflected images in both fields of the second pattern made with the same design on the same sample. And
A dark field and bright field illumination system that illuminates the same point of the first pattern;
A dark field image detector for detecting a reflected image of the first pattern in the dark field;
A bright field image detector for detecting a reflected image of the first pattern in the bright field;
A dark field comparator coupled to the dark field image detector to form a dark field difference signal compared to a reflected image in the dark field of the second pattern;
A bright field comparator coupled with the bright field image detector to form a bright field difference signal in comparison with a reflected image in the bright field of the second pattern;
Said dark-light, combined with each of the comparators of each view, the dark-light, both differential signal and data processing and a data processor to form a defect list centrally first pattern
Data processing of the dark and both differential signals
1) generating a two-dimensional histogram based on the difference signals of the dark and bright visual fields;
2) identifying a region in the two-dimensional histogram associated with a process or surface variation of the sample; and
3) excluding such regions from the combination of data representing a two-dimensional histogram; and
4) determining a defect list of the first pattern based on data from the rest of the data combination and not based on data from the excluded area of the two-dimensional histogram.
An automatic defect inspection apparatus characterized by that.   After the automatic defect inspection device is coupled with the data processor and post-processes the defect list of the first pattern, after selecting and eliminating only known and harmless false defects therein The automatic defect inspection apparatus according to claim 13, further comprising a processing processor.   The automatic defect inspection apparatus according to claim 13, wherein the dark field and bright field illumination system includes a dark field illumination subsystem and a bright field illumination subsystem.   The automatic defect inspection apparatus according to claim 15, wherein the dark-field and bright-field illumination subsystems emit light having different frequencies. The dark field and bright field illumination subsystem provides narrowband illumination light;
17. The automatic defect inspection apparatus according to claim 16, wherein the dark field image detector includes a spatial filter and allows dark field reflected light to pass therethrough to enhance defect detection capability. The sample is transparent, see the darkness of the second pattern, known transmission images in each field of view,
The automatic defect inspection apparatus includes:
A dark field transmitted light detector for detecting a dark field transmitted image obtained from the first pattern;
A bright-field transmitted light detector for detecting a bright-field transmitted image;
A dark field transmitted light signal comparator coupled to the dark field transmitted light detector to generate a differential signal by comparing dark field transmitted light signals for the same observation point of the first pattern and the second pattern;
A bright field transmitted light signal comparator, coupled to the bright field transmitted light detector, for comparing the bright field transmitted light signals for the same observation point of the first pattern and the second pattern with each other to generate a differential signal; The processor is combined with the transmitted light signal comparators of the dark and bright fields, and processes the dark and dark differential signals to centrally form a defect list of the first pattern. The automatic defect inspection apparatus according to claim 13.   The automatic defect inspection apparatus includes a post-processing processor that is coupled to the data processor, receives a first defect list, and selects and removes only data of known harmless false defects. 18. An automatic defect inspection apparatus according to 18.   The automatic defect inspection apparatus according to claim 18, wherein the dark field and bright field illumination system includes a dark field illumination subsystem and a bright field illumination subsystem.   21. The automatic defect inspection apparatus according to claim 20, wherein the dark field and bright field illumination subsystem emit light having different frequencies. The dark field illumination subsystem provides narrowband illumination light;
The dark field reflected image detector has a first spatial filter to pass dark field reflected light and enhance the detection capability;
The automatic defect inspection apparatus according to claim 21, wherein the dark field transmission image detector includes a second spatial filter, and allows dark field transmission light to pass therethrough to enhance detection capability. With respect to the same observation point with illumination systems having different frequencies, the second pattern is arranged to refer to a plurality of known dark field reflection images,
The dark field and bright field illumination systems are set to irradiate the same observation point of the sample with illumination systems having different frequencies.
The dark field image detector is composed of a plurality of separate detectors and is set to detect a dark field reflected image of a first pattern based on multiple light sources,
The dark field comparator is coupled to each of the dark field image detectors, compares the dark field reflected image of the plurality of first patterns with the dark field reflected image of the second pattern, and outputs a difference signal thereof. The automatic defect inspection apparatus according to claim 13, wherein the automatic defect inspection apparatus is formed. With respect to the same observation point with illumination systems having different frequencies, the second pattern is arranged so that a plurality of known bright-field reflection images are referenced,
The dark field and bright field illumination systems are set to irradiate the same observation point of the sample with illumination systems having different frequencies.
The bright-field image detector is composed of a plurality of separate detectors and is arranged to detect a bright-field reflected image of a first pattern based on multiple light sources, respectively.
The bright field comparator is combined with each of the bright field image detectors, and the obtained bright field reflected images are compared with the bright pattern reflected images of the second pattern, respectively, to form a difference signal thereof. The automatic defect inspection apparatus according to claim 13. The dark field and bright field illumination systems are:
A single laser illumination source;
A beam splitter installed to receive illumination light traveling in one direction from the laser light source and reflect it;
Including a condenser lens for guiding the light beam to an observation point on the sample,
The dark field image detector is installed at a low elevation angle to receive reflected light from the sample,
The bright-field light detector is disposed facing the observation point so as to receive a bright-field image incident through the condenser lens and a beam splitter after being reflected from the sample. Item 14. An automatic defect inspection apparatus according to Item 13. The dark field and bright field illumination systems are:
A narrow-band laser light source configured with a constant-frequency narrow-band laser and placed at a low elevation with respect to the sample to provide a dark field illumination system;
A mercury arc lamp,
A first condenser lens that passes light from the mercury arc lamp;
A first beam splitter installed to reflect a light beam passing through the first condenser lens in one direction;
A second condenser lens that directs the bright field light from the first beam splitter to a common observation point with the dark field illumination on the sample;
The dark field image detector comprises:
The second beam splitter is arranged to receive the reflected light from the sample via the second condenser lens and the first beam splitter, and in particular, by applying a dichroic dichroic film selectively. However, in order to reflect only the dark-field illumination light emitted from the laser, the set angle is mutually adjusted with the positional relationship between the second condenser lens and the first beam splitter. A second beam splitter installed
A third lens having a function of focusing light of a dark field image reflected from the second beam splitter;
A dark field detector for detecting the dark field reflected image;
The second condenser lens is provided on the same line as the first and second beam splitters and behind the second beam splitter, and passes through the light beam carrying the remaining bright-field reflected image, that is, through the second beam splitter. A fourth lens arranged to receive and narrow the incoming rays,
14. The automatic defect inspection apparatus according to claim 13, further comprising a bright field photodetector installed behind the fourth lens and optically coupled thereto for receiving the bright field reflected image. The sample is transparent;
The automatic defect inspection apparatus includes:
When transmitting light to the sample, a dark-field transmitted light detector installed at a low elevation angle with respect to the sample surface to receive a dark-field transmitted image from the sample;
26. The automatic defect inspection apparatus according to claim 25, further comprising a bright field transmitted light detector disposed on the rear surface of the sample for receiving a bright field transmitted image. The sample is transparent and is arranged to transmit light;
The automatic defect inspection apparatus includes:
A fifth condenser lens arranged behind the sample to expand the transmitted light;
A dark field transmitted light image detector;
A bright field transmitted light image detector,
The dark field transmitted light image detector is:
A third beam splitter installed behind the fifth condenser lens, which receives the transmitted light from the sample, and its reflective surface is composed of a dichroic dichroic film to selectively transmit only the dark field transmitted light. The second beam splitter is set to reflect, while allowing all other rays to pass through. The second beam splitter transmits the dark field transmitted light in a direction different from the optical path defined by the fifth condenser lens. The third beam splitter set to reflect
A sixth lens set to squeeze and collect the dark field transmitted light reflected from the third beam splitter at the focal point;
A dark field transmitted light detector provided to receive a dark field transmitted image,
The bright field transmission image detector is:
A seventh lens, which is placed behind the third beam splitter and aligned with the fifth condenser lens so that all the remaining rays, i.e. the light received by the third beam splitter, is received. A seventh lens arranged to narrow the field of transmitted light,
27. The automatic defect inspection apparatus according to claim 26, further comprising: a bright-field transmitted light detector which is set immediately behind the seventh lens and receives a bright-field transmitted image from a sample. A method of comparing a second pattern on the same sample made with the same design when searching for a defect present in a first pattern on a sample, wherein the second pattern is searched at least once In an automatic defect inspection method having first and second responses known in
a) performing a test search on at least two corresponding identical points of the first pattern of samples to obtain at least two responses;
b) processing the first response of the first pattern with a detector;
c) processing the second response of the first pattern with a detector;
d) comparing the first response of step b) with the first response at the same point of the second pattern to generate their differential signal;
e) comparing the second response of step c) with the second response at the same point of the second pattern to generate their difference signal; and f) the difference signal of the first response. two of the difference signal of the second response to the data processing comprises a step of determining a defect list centrally first pattern,
The step of data processing of the difference signal in the first response and the difference signal in the second response is as follows.
1) generating a two-dimensional histogram based on the difference signal in the first response and the difference signal in the second response;
2) identifying a region in the two-dimensional histogram associated with a process or surface variation of the sample; and
3) excluding such regions from the combination of data representing a two-dimensional histogram; and
4) Based on the data from the remaining combination of the data, and not based on the data from the excluded area of the two-dimensional histogram, comprising the steps that determine the defect list of the first pattern,
An automatic defect inspection method characterized by that. For the purpose of searching for defects present in the first pattern on the sample, the same observation points of the second pattern on the same sample made with the same design are compared, and the second pattern is known by at least one search. In an automatic defect inspection apparatus deployed to generate first and second responses,
At least one image search mechanism for observing and searching for observation points of the first pattern on the sample to obtain at least two responses;
A first response detector for detecting a first response;
A second response detector for detecting a second response;
A first comparator coupled to the first response detector and generating a first differential signal of the response by comparing the output of the first response detector and the output of the first response in the second pattern;
A second comparator coupled to the second response detector and generating a second differential signal of the response by comparing the output of the second response detector and the output of the second response in the second pattern;
A processor coupled to the first second comparator for processing the first and second differential signals to centrally determine one defect list of the first pattern ;
Note that processing the first and second differential signals includes:
1) generating a two-dimensional histogram based on the first and second difference signals;
2) identifying a region in the two-dimensional histogram associated with a process or surface variation of the sample; and
3) excluding such regions from the combination of data representing a two-dimensional histogram; and
4) determining a defect list for the pattern based on data from the remainder of the data combination and not based on data from the excluded region of the two-dimensional histogram.
An automatic defect inspection apparatus characterized by that. The method of claim 1, wherein the two-dimensional histogram is generated substantially without first thresholding the difference signal. The method of claim 1, further comprising: generating a histogram greater than two dimensions based on data derived from greater than two channels and processing the histogram to exclude regions associated with process or surface variations of the sample. the method of. The method of claim 1, wherein the excluded region represents a composite function of both the dark field and bright field difference signals. The method of claim 1, wherein the bright field reflected image is detected using a time delay integrated (TDI) sensor.

Images