JP3410847B2 - Pattern defect inspection method and inspection device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern defect inspection technique, and more particularly to a pattern defect inspection method and an inspection apparatus for inspecting a pattern defect formed on a sample such as a photomask used in semiconductor manufacturing.

[0002]

2. Description of the Related Art One of the major causes of a decrease in yield in the manufacture of large scale integrated circuits (LSIs) is a defect occurring in a photomask or reticle used when manufacturing a device by a photolithography technique. To be Therefore, an apparatus for inspecting such defects has been actively developed and put into practical use.

The conventional mask defect inspecting apparatus is roughly divided into two methods, that is, two chips on which the same pattern is drawn are observed by different detecting means, and the difference between them is compared and detected by an appropriate defect detecting means. There is a method of detecting a defect by observing a chip on which a pattern is drawn by a detecting means and comparing this with design data of the pattern by an appropriate defect detecting means. In the former case, since two chips having the same pattern are observed respectively, if the same defect exists, there is a drawback that the defective part cannot be detected. In the latter case, such a problem does not occur because the design data is compared.

As an apparatus for inspecting using design data, for example, an apparatus shown in a document (high-precision fully automatic reticle inspection apparatus for VLSI, electronic material, September 1983, p47), or Japanese Patent Publication No. 1-40489. The technology disclosed in the publication is cited. It is desirable that the design data used when manufacturing (drawing) the reticle and the design data used when performing inspection by inputting to the inspection device match. An efficient system can be constructed by using the described drawing and inspection devices in pairs.

Most conventional masks to be inspected are formed by vapor-depositing non-transparent chrome on transparent glass, drawing a pattern, and then etching the mask. Optically, the transmittance is almost 100%. The characteristic was almost 0%. On the other hand, since the design data is originally binary data of 1 and 0, it was relatively easy to associate the measurement data obtained by the sensor element through the optical system with the design data.

By the way, in recent years, various phase shift masks have been devised and put to practical use in order to improve the resolution characteristics of the exposure apparatus as much as possible. Among them, a mask called a halftone mask is regarded as promising because it is easy to design a pattern. This halftone mask uses a semitransparent film such as silicon nitride instead of the conventional chromium film.

With this type of mask as well, it is necessary to carry out a defect inspection of the pattern formed of the halftone film as in the conventional chromium mask. However, the transmissivity of the semitransparent film is about 10% to 70% in some cases, and the optical contrast is poor. Therefore, it is the current situation that the halftone mask cannot sufficiently detect defects.

[0008]

As described above, in the conventional pattern defect inspection, it is difficult to detect minute defects with high accuracy in a mask made of a semitransparent film such as a halftone mask. .

The present invention has been made in consideration of the above circumstances, and it is an object of the present invention to provide a mask made of a semi-transparent film such as a halftone mask with minute defects like the conventional chromium mask. An object of the present invention is to provide a pattern defect inspection method and an inspection apparatus capable of accurately detecting

[0010]

In order to solve the above problems, the present invention employs the following configurations. That is, the book
The invention irradiates a sample having a pattern with light of a predetermined wavelength, receives a pattern image of the sample by a light receiving element through an optical system, and multivalued design pattern image from binary design data corresponding to the pattern image. In a pattern defect inspection method for inspecting a defect on a sample surface by creating data and comparing the observed data corresponding to the received pattern image with the design pattern image data, the observation data and the design pattern image data are combined. Before comparison, in order to match the observation data of the semi-transparent portion where the semi-transparent portion of the pattern on the sample surface is regarded as the non-transparent portion with the design pattern image data corresponding to the semi-transparent portion,
And adjust the offset of the photodetector amplifier and the observation data of the transparent portion of the sample surface on the pattern to match a design pattern image data corresponding to the transparent portion, after adjusting the gain of the light receiving element amplifier, the translucent portion By the defect of
A negative signal whose signal level is lower than that of the translucent part
Make sure that the light receiving element amplifier has a predetermined
The amount of offset is given and the design pattern image is
The feature is that the same offset amount is given to the data.

Here, the following are preferred embodiments of the present invention. (1) The sample must be a photomask or reticle using a phase shifter, especially a halftone mask. (2) The sample is provided with a pattern of semi-transparent areas and transparent areas above a certain area, and when adjusting the offset and gain of the light-receiving element amplifier, the semi-transparent part is first illuminated and the light-receiving element obtained at that time is obtained. Adjust the offset of the light receiving element amplifier so that the amplifier output becomes zero, then illuminate the transparent part and adjust the gain of the light receiving element amplifier so that the light receiving element amplifier output obtained at that time becomes a predetermined value. thing. (3) After adjusting the gain of the light receiving element amplifier, use the data obtained by adding a predetermined offset amount to the output of the light receiving element amplifier as the observation data, and compare this observation data with the design pattern image data. (4) The predetermined offset amount is larger than the data output change amount caused by the phase change at the semitransparent pattern edge. (5) A predetermined offset amount corresponding to the offset of the amplifier output is added to the design pattern image data.

Instead of comparing the observation data with the design pattern image data, the two pattern images of the sample are received by the two light receiving elements through the two optical systems, and the received pattern images are converted into the respective pattern images. By comparing the two corresponding observation data,
It is also conceivable to detect and inspect a pattern defect on the sample surface and to adjust the offset and gain of the light receiving element amplifier at this time as in the above case.

Further, the present invention is a pattern defect inspection apparatus used for defect inspection of a pattern having a semi-transparent film such as a halftone mask, and means for irradiating a sample having a pattern with light of a predetermined wavelength, A light receiving element that receives a pattern image of the sample obtained by light irradiation by this means through an optical system, a light receiving element amplifier that amplifies the output of the light receiving element, and 2 corresponding to the pattern image of the sample.
Means for generating multi-valued design pattern image data from valued design data, and design pattern image data corresponding to the semi-transparent part of observation data of the semi-transparent part in which the semi-transparent part of the pattern on the sample surface is regarded as the non-transparent part so as to coincide with a means for adjusting the offset of the light receiving element amplifier, the observation data of the transparent portion of the sample on the surface pattern to match with the design pattern image data corresponding to the transparent portion, of the light receiving element amplifier Means to adjust the gain,
Semi-transparent with the offset and gain adjusted
Signal level lower than semi-transparent areas caused by defect in area
The negative signal to prevent it from being cut off at zero level.
A predetermined offset amount is given to the optical element amplifier and
The same offset amount is applied to the design pattern image data.
Means for giving, the offset and gain are adjusted ,
Further, it is provided with a means for inspecting a pattern defect on the sample surface by comparing the observation data obtained from the output of the light receiving element amplifier with the design pattern image data in the state where a predetermined offset amount is given. It is characterized by

[0014]

The preferred embodiments of the present invention are as follows. (1) The sample must be a photomask or reticle using a phase shifter, especially a halftone mask. (2) If the optical characteristics of the pattern on the sample surface are semi-transparent to the irradiated light, the conversion function that creates multi-valued design pattern image data from binary design data has at least three peaks. Be a function with. (3) Applying a two-dimensional digital filter to the design data as a means of creating a multi-valued design pattern image from the binary design data. (4) As a means of creating a multi-valued design pattern image from binary design data, a predetermined offset value is superimposed (added / subtracted) on the multi-valued design pattern image data.
thing. (5) Use a limiter means for limiting the amplitude at a predetermined signal level as a means for producing a multi-valued design pattern image from binary design data. (6) Adjust the offset and gain of the observation data output according to the difference in the optical characteristics of the pattern on the sample surface, and after filtering the observation data, compare the design pattern image data with the observation data after the filtering. Then
Detecting pattern defects on the sample surface. (7) If the optical characteristics of the pattern on the sample surface are semi-transparent to the irradiated light, the filtering process should be a low-pass filter that reduces high-frequency components. (8) The offset and gain of the observation data output are adjusted according to the difference in the optical characteristics of the pattern on the sample surface, the observation data is filtered, and the binary data is adjusted so that it matches the observation data after filtering. A multi-valued design pattern image is created by changing a conversion function when a multi-valued design pattern image is created from design data, and the design pattern image data and the observed data are compared to detect a pattern defect on the sample surface. . (9) Observe a predetermined test pattern on the sample surface before or during inspection, and estimate the optimum conversion function or the optimum filtering coefficient from the observed profile,
Select automatically.

[0016]

[0017]

According to the present invention, when the halftone mask is inspected in the inspection method by the design data comparison method, the zero point of the output of the light receiving element amplifier is treated so that the semitransparent portion of the pattern on the sample surface is regarded as the nontransparent portion before the inspection. Calibration processing for adjustment and gain adjustment was performed. In other words, first illuminate the semi-transparent part that was provided in advance, adjust the zero point of the photo-receiver amplifier output so that the photo-receiver amplifier output obtained at that time is zero, and then adjust the transparent part that was provided in advance. The gain of the light-receiving element amplifier output can be adjusted so that the light-receiving element amplifier output obtained at the time of illumination becomes the gain reference value.

As an inspection device, an offset adjustment margin of the light-receiving element amplifier is set so that the output of the light-receiving element amplifier obtained when the semi-transparent portion is illuminated can be lowered to zero. The gain of the light receiving element amplifier is set so that the output of the light receiving element amplifier can be raised to the gain reference value.

Next, a predetermined offset amount is added.
The maximum value and the minimum value of the output obtained as a result are set as the high level reference value and the low level reference value, respectively. The low-level reference value and the high-level reference value are determined in accordance with the level of design data which will be the inspection comparison standard described below. That is, 2
Match the low-level value and the high-level value of the multi-valued design data based on the value data. That is, the semi-transparent part of the pattern is regarded as a non-transparent part, and the sensor amplifier output is made to match the output value of the design data corresponding to the non-transparent part. Try to match the output value.

With such a configuration, by performing the calibration process before performing the inspection, the semitransparent portion of the pattern on the sample surface of the halftone mask can be regarded as the nontransparent portion, and the observation data It is possible to make it almost coincide with the design data. As a result, it is possible to eliminate the inconsistency in the data between the observation data and the design data that has been conventionally obtained with the halftone mask, and even if the two are compared, a comparative inspection can be performed without generating a pseudo defect.

Further, in the present invention, in the case of inspecting the halftone mask in the inspection method by the design data comparison method, the reason why the detection sensitivity is not increased as much as the chrome mask due to the frequent occurrence of pseudo defects is the pattern edge. Focusing on the output characteristics of the data, the appropriate filtering process is performed to match the output characteristics of the pattern edge of the design data and the observed data to prevent the occurrence of spurious defects to the minimum. There is.

For this reason, in the present invention, firstly, the design data is subjected to a convolution integration of a function in which profile characteristics peculiar to the halftone film are convoluted, and secondly, the observation data is subjected to a filtering process for smoothing a sharp data change. I am trying. That is, the essence of the present invention is to add an appropriate filter means to the design data or the observation data in order to match the output characteristics at the pattern edge portion of the design data and the observation data according to the characteristics of the mask to be inspected. There is.

With such a configuration, even when the halftone mask is inspected, the output profiles of the pattern edges of the design data and the observation data can be well matched, and the occurrence of pseudo defects can be minimized. This can be prevented as much as possible, and the detection sensitivity can be increased. At that time, the conventional comparison algorithm can be used as it is. Further, even halftone masks having different characteristics can be dealt with by selecting a filter according to the characteristics, and it is possible to provide a more practical defect inspection apparatus.

[0024]

Embodiments of the present invention will be described below with reference to the drawings. (Embodiment 1) As an example, a mask defect inspection apparatus for detecting a defect in a mask pattern using design data will be described. The basic structure of such a mask inspection is shown in a document (high-precision fully automatic reticle inspection device for VLSI, electronic material, September 1983, p47). The inspection method is shown in FIG. 1 and the device configuration is shown in FIG. Here is an example:

As shown in FIG. 1, the inspection is carried out by enlarging the mask pattern using an optical system or the like and observing a thin strip-shaped portion of W = 500 μm continuously (actually, the table continuously moves). To be executed.

As shown in FIG. 2, a mask 2 to be inspected is placed on an XYθ table 1, and a pattern is irradiated by an appropriate light source 3 and illumination lens 13. A pattern image is formed on the photodiode array (light receiving element) 5 using the objective lens 4, and the observation data A / D converted by the sensor circuit 6 including the sensor amplifier unit and the A / D conversion unit is obtained. This observation data is sent to the data comparison circuit 8 together with the position data from the position circuit 7. On the other hand, the pattern design data is sent from the magnetic disk device 9 to the bit pattern generation circuit 11 through the control computer 10, and the graphic data is binarized and sent to the data conversion circuit 21.

In the data conversion circuit 21, a point spread function equivalent to the optical system is convolutively integrated with the binary bit pattern data to create multivalued design pattern image data. This is because the observation data is in a state in which the low-pass filter on the spatial frequency acts due to the resolution characteristic of the objective lens 4 and the aperture effect of the photodiode array 5, and thus the design data is ideally filtered. This is to simulate various observation data.

The data comparing circuit 8 compares the two in accordance with an appropriate algorithm, and a portion where the design pattern image data and the observation data do not match is determined to be a defect. Examples of this comparison algorithm include a method of directly comparing data, and a method of obtaining and comparing a differential value (gradient vector) of pattern data as disclosed in JP-A-62-266406.

In such an inspection method, as shown in FIG.
The chrome mask pattern is actually measured as shown in FIG. Then, the output of the light receiving element has a waveform with a blunt edge of the pattern as shown in FIG. 3B due to the resolution characteristic of the objective lens 4 and the aperture effect of the photodiode array 5 as described above. . On the other hand, FIG.
As the design pattern image data obtained by performing the filtering process on the design data as shown in Fig. 4, a waveform similar to the observation data is obtained as shown in Fig. 4B. Therefore, minute defects can be detected with high sensitivity by comparing the design pattern image data with the measurement data according to an appropriate algorithm. As a method of this filter processing, for example, a method disclosed in Japanese Examined Patent Publication No. 63-3450 can be used.

On the other hand, when the halftone mask is inspected as shown in FIG. 5A, the output of the light receiving element is an offset Vh at the semi-transparent portion as shown in FIG. 5B.
Has a waveform like that of, and does not match the design pattern image data shown in FIG. For this reason, if defect detection is performed as in the conventional case, spurious defects often occur due to data inconsistency, even though there is no defect, and comparison inspection cannot be performed.

Therefore, in the present embodiment, when inspecting a halftone mask, the halftone calibration process described below is performed before the inspection is performed, thereby suppressing the occurrence of pseudo defects and enabling an accurate inspection. There is.

First, as a precondition, as shown in FIG.
The halftone mask to be inspected is provided with a semitransparent area A and a transparent area B in advance. The semi-transparent area A and the transparent area B may be outside the inspection area as shown in FIG. 6A or may be inside the inspection area as shown in FIG. 6B. Both sizes are set to be sufficiently larger than the size of the light receiving element when projected onto the light receiving element through the area on the mask.

When the mask to be inspected cannot be provided with such a region, a calibration mask provided with a semitransparent region and a transparent region may be prepared in advance. In that case, the characteristics of the halftone film of the calibration mask are made to match those of the mask to be inspected.
Layout coordinates on the mask (Xh, Y
h) and the layout coordinates on the mask of the transparent area B (Xt, Y
t) is determined in advance.

Normally, the sensor amplifier is adjusted so that the output is 0 at the non-transmissive portion where the light does not strike and the output becomes the maximum value Vmax at the transmissive portion where the light strikes the maximum, or as close as possible. Therefore, in the state where the calibration process is not performed, the sensor amplifier output obtained through the semitransparent portion such as the halftone film has a value Vh which is an intermediate value between 0 and Vmax, and the output at the pattern edge portion of the halftone film is as shown in FIG. As shown in (b).

Therefore, the following calibration process is performed. First, the XYθ table is moved so that the semi-transparent portion is illuminated and the light can be projected through the area. At that time, the stop position of the table may be determined based on the layout coordinates (Xh, Yh). As shown in FIG. 7A, the sensor amplifier output is adjusted so that the sensor amplifier output obtained at that time becomes zero. As the inspection device, a sufficient adjustment margin of the offset of the sensor amplifier may be provided so that the sensor amplifier output obtained when the semitransparent portion is illuminated can be reduced to zero.

Next, the transparent portion is illuminated and the table is moved so that the transparent portion can be projected onto the light receiving element. As shown in FIG. 7B, the gain on the sensor amplifier side is adjusted so that the sensor amplifier output obtained at that time becomes the preset gain reference value G. As the inspection device, a gain margin of the sensor amplifier may be set so that the sensor amplifier output obtained when the transparent portion is illuminated is increased to the gain reference value G.

Further, a constant offset (V0) is added to the output of the amplifier to obtain observation data. The circuit configuration in this case is shown in FIG. As a result, the observation data obtained through the semitransparent pattern edge portion after the calibration process is as shown in FIG. 9B. Here, the output level from the semi-transparent portion of the observation data is the low level reference value (V0), and the output level from the transparent portion is the high level reference value (V1).

The low-level reference value and the high-level reference value are determined in accordance with the level of design data which will be the inspection / comparison reference described below. That is, by performing the convolution integration (sum of products operation) performed by the data conversion circuit 21 described in FIG.
The design data at the pattern edge portion as shown in 0 (a) becomes a multi-valued waveform (design pattern image data) reflecting the blur of the optical system as shown in FIG. 10 (b). The method of this convolutional integration is, for example, Japanese Patent Publication 1-40.
The method disclosed in Japanese Patent No. 489 may be used.

Here, the low level value V0 and the high level value V1 of the design pattern image data are coefficient and offset addition processing at the time of calculation, and are converted into the low level reference value V0 and the high level reference value V1 of the sensor amplifier output. Can be matched. That is, the semi-transparent part of the pattern is regarded as a non-transparent part, the output of the sensor amplifier is made to match the output value of the design pattern image data corresponding to the non-transparent part, and the transparent part is directly designed so that the sensor amplifier output corresponds to the transparent part. Match the output value of the pattern image data. As a result,
For example, the sensor output at the edge portion of the halftone pattern and the output value of the design data are as shown in FIGS. 9B and 10B, and they can be substantially matched.

When the inspection is executed after the above halftone calibration process is executed, the semitransparent portion of the pattern can be regarded as the nontransparent portion, and the sensor output of the halftone pattern and the output of the design data should be matched. You can As a result, as shown in FIG. 11, if there is a defect in the halftone pattern, a difference occurs between both the measurement data and the design pattern image data. Can be detected well. As a comparison algorithm at that time, a method of directly comparing data or a method of obtaining a differential value (gradient vector) of data and comparing the data may be used.

The low-level reference value V0 is usually set to 0 in order to gain the dynamic range of the signal and to easily adjust the gain of the amplifier. However, the inventors have found out that a new effect can be obtained by adding a certain value. That is, as shown in FIG. 12A, the waveform of the pinhole-shaped defect signal of the halftone pattern is a positive signal (more than the halftone level) obtained by the light passing through the pinhole when the size is relatively large. The signal that appears on the plus side) is large, but as shown in FIG. 12B, when the size is as small as 1 μm or less, the positive signal obtained by the light passing through the pinhole is small. On the other hand, the waveform at the edge portion of the defect appears on the minus side (negative signal) of the halftone level due to the phase shift effect, and becomes larger than the signal appearing on the plus side.

Therefore, the observed data when V0 = 0 is only the positive signal as shown in FIG. 13 (a), and it becomes difficult to detect the minute defect. However, the value d is larger than the signal amplitude appearing on the minus side during calibration.
When a constant value is added (V0 = d), the negative signal at the edge portion is stored as it is, as shown in FIG.

On the other hand, for the design data, for example, if the structure of the data conversion circuit 21 is composed of a product-sum calculation means and an offset superposition means as shown in FIG.
A waveform obtained by applying a point spread function as shown in FIG. 15 (b) from the design data as shown in FIG. 15 and a waveform with an offset superimposed as shown in FIG. 15 (c) are obtained. Therefore,
Comparing FIG. 13 (b) and FIG. 15 (c), the negative signal at the defect edge portion appears as the difference between the two as it is, so that it becomes possible to detect a defect of a size that could not be detected until now. The detection performance is further improved.

The present invention is not limited to the above embodiment. The embodiment relates to a mask defect inspection apparatus usually called a database comparison type that detects a defect of a mask pattern using design data.
A mask defect inspection apparatus called a normal die-to-die comparison type in which two chips on which the same pattern is drawn is observed by different detection means and the difference between the two chips is compared and detected by an appropriate defect detection means may be used. In this case, there are two sensor amplifiers from which observation data can be obtained, but the same effect can be obtained by performing the above-described calibration process on the outputs of the two amplifiers. In addition, various modifications can be made without departing from the scope of the present invention. (Second Embodiment) Next, a second embodiment of the present invention will be described.
The basic structure of the mask defect inspection apparatus is the same as that of the first embodiment, and its detailed description is omitted.

When the halftone mask is inspected as shown in FIG. 16 (a), the output of the photodiode array 5 has a waveform in which the offset is superimposed as shown in FIG. In some cases, it does not match the design data shown in (b), pseudo defects frequently occur, and the detection sensitivity cannot be increased as much as that of the chromium mask.

As a result of diligent research on this point, the present inventors have found that there is a drop in the output at the pattern edge portion (A portion in FIG. 16B) in the halftone mask, which causes the observation data profile to be the objective. It was found that the characteristics were different from those due to the resolution characteristics of the lens 4 and the aperture effect of the photodiode array 5. The second embodiment and the third embodiment described later solve the above problems.

In the second embodiment, a function incorporating the profile characteristics peculiar to the halftone film is used when the design data is filtered in the above-described data conversion circuit 21 to simulate ideal observation data. Performs convolution product sum calculation.

This function is estimated and obtained as a point spread function for the halftone film from the edge profile of the halftone film and the glass portion (FIG. 16). In addition to this, as the estimation method, means such as observing various lattice patterns at predetermined intervals, obtaining spatial frequency characteristics, and performing Fourier transform can be considered.

Regardless of which method is used, the point spread function for the halftone mask will exhibit a signed characteristic (including a negative coefficient term) as shown in FIG. When the convolution product sum operation of the point spread function for the halftone mask is performed by the data conversion circuit 21 from the design data as shown in FIG. 18A, the expected value of the observed data as shown in FIG. Become. Here, in order to bring it closer to the actual observation data, the offset amount is adjusted, the signal amplitude is adjusted, and the signal limiter is operated.

The offset means and the signal amplitude adjusting means are arranged in accordance with the waveform of the observation data shown in FIG.
The amount of light corresponding to the semitransparent portion of the halftone film is added as an offset to the waveform of (b), and the amplitude is adjusted so that the amount of light corresponding to the glass portion matches. As a result, the waveform shown in FIG. 18C is obtained.

Limiting means is provided for flattening the waveform C portion shown in FIG. 18 (b). In the actual observation data of the halftone mask, overshoot does not occur in the C portion as shown in FIG. 16B, and it changes gently. For this reason, it may be necessary for the limiting means to not only cut at the threshold value but also to cut at the threshold value and to operate a small low-pass filter on the spatial frequency. The presence or absence of this overshoot is considered to be due to the saturation of photoelectric conversion in the photodiode array, etc., and the limiter may not be operated depending on the situation.

That is, the configuration in the data conversion circuit 21 is as follows.
As shown in FIG. 19, it comprises a product-sum calculation means, an offset superposition means, an amplitude adjustment means, and a limit means.
However, generally, by selecting an appropriate coefficient in the product-sum calculation, superimposing an offset and adjusting the amplitude can be collectively processed.

The configuration of the data conversion circuit 21 is such that a mask is prepared by preparing a sample of such an input / output function in a table format in advance and inspecting it by a simulation without using the means composed of the above-mentioned constituent elements. Even if the optimum one is selected and used, the same effect can be obtained.

In any of these means, since the profile characteristics peculiar to the halftone film are reflected in the reference data obtained from the design data, the design data to be compared is shown in FIG.
As shown in FIG. 8 (d), the value is close to the profile of the observation data, and as a result, it is possible to prevent the occurrence of spurious defects even in the halftone mask to the minimum, and it is possible to increase the detection sensitivity and to reduce the size. It is possible to detect even defects. (Third Embodiment) A third embodiment is a method in which the observation data is filtered, the edge portion is made smooth, and the comparison in the comparison circuit 8 is performed for the chrome mask. In the observation data of the conventional halftone mask, as shown in FIG. 16C, the waveform from which the offset is removed has a sharp rise in the B portion.

Here, the added filter circuit performs a filtering process for smoothing this waveform, and further amplifies the signal amplitude to the same level as the chrome mask. Then, the data comparison circuit 8 compares with the inspection reference data (design pattern image data) for the chrome mask shown in FIG.

This filter circuit is disclosed in, for example, Japanese Patent Laid-Open No. 03-
A digital filter as disclosed in Japanese Patent No. 124052 may be used. Contrary to the second embodiment, the characteristics of the filter may be obtained by obtaining the frequency characteristic so that the edge portion becomes gentle, and further performing the Fourier transform to obtain the filter function.

To properly select the characteristics of this filter, it is possible to input the waveform shown in FIG. 16 (b) instead of the waveform shown in FIG. 16 (c) as the observation data processed by the filter means. In other words, the same effect can be obtained by reversing the order of the observation data offset / amplitude adjusting means and the filtering means.

Similar to the second embodiment, this filter means is prepared in advance with a sample of such a function, and the optimum one suitable for the mask to be inspected in the simulation is selected and used. Good. By doing this, the observation data after calibration is shown in FIG.
0 (a), and the observed data after further filtering has a smooth undershoot and a sharp change near the edge as shown in FIG. 20 (b), and FIG. The value is close to the profile of the design data. As a result, even with a halftone mask, the occurrence of pseudo defects can be prevented to a minimum, the detection sensitivity can be increased, and even minute defects can be detected. (Embodiment 4) As a fourth embodiment of the present invention, it is conceivable that the processing in the data conversion circuit 21 is changed and the comparison data in the comparison circuit 8 is performed in another format. This is because the observation data is subjected to a calibration process such that the offset is reduced and the dynamic range S is expanded to S ′ so that the observation data value of the semitransparent portion of the halftone film becomes zero as shown in FIG. 16C. This is a method in which the inspection reference data derived from the design data is brought close to this in the applied state.

The process is as shown in FIG. That is, the design data shown in FIG. 21A is multiplied by an appropriate point distribution function by the product-sum calculation means of the data conversion circuit 21,
The data shown in (b) is obtained. Furthermore, B in FIG.
In order to have a sharp change in consideration of the portion, a slight offset subtraction is performed to obtain the data shown in FIG. 16 (c). Then, by adjusting the amplitude so as to compensate for this offset subtraction portion, the data shown in FIG. 21D is obtained. The observation data is processed as shown in FIG. 16 (c) as in the conventional example.

With the above measures, the comparison in the comparison circuit
As shown in FIGS. 16 (c) and 21 (d), it is possible to perform data matching with each other in which the rising edges near the edges are sharply matched. As a result, even with a halftone mask, the occurrence of pseudo defects can be prevented to a minimum, the detection sensitivity can be increased, and even minute defects can be detected.

As described above, as in the second and third embodiments, the signal correction processing is performed by either the observation data input to the data comparison circuit 8 or the design pattern image data. Is also effective, and both may be used together, as described in the fourth embodiment. The algorithm for the comparison inspection at that time may be the conventional one.
Further, even if there are halftone masks having different characteristics, if the control computer 19 controls so that the optimum filter can be selected each time according to the halftone masks, it is possible to deal with it, and the practicality is high.

The function or filter prepared in advance is
You may make it represented by several coefficients and can make it selectable or set arbitrarily. These selections and settings may be performed locally in the data conversion circuit 21 or the filter circuit 22, or may be performed through the control computer 19. In the latter case, the optimum function or filter coefficient may be obtained.

Further, as the filter means, the offset means, etc., a product-sum operation circuit and a numerical operation circuit according to the definition are prepared, and a function ROM () having some variations is calculated in advance by means of simulation or the like. Read Onl
y Memory) and install it in the device, and the circuit just reads this ROM in a table lookup system,
It is also possible to have the same effect.

[0064]

As described in detail above , according to the present invention, in the inspection of the halftone mask, the semi-transparent portion of the pattern on the sample surface is regarded as the non-transparent portion before the inspection is performed,
By performing the calibration process for adjusting the offset and the gain of the sensor amplifier output, it is possible to make the observation data and the design data substantially coincide with each other. As a result, by comparing the observed data with the design data, it is possible to perform a comparative inspection with the same detection algorithm as the conventional one, and it is possible to find a defect that could not be found in the conventional halftone mask. became.

As a result, even if a halftone mask is used in the manufacture of an LSI, the shape defect of the pattern formed by halftone can be detected with high accuracy, and the reduction of the yield in the photolithography process can be prevented. . Therefore, the cost of manufacturing the LSI can be reduced, and the effect is large.

Further , according to the present invention, in the inspection of the halftone mask, an appropriate filter processing is performed so as to make the output characteristics of the design data and the observation data at the pattern edge portion coincide with each other, thereby generating a pseudo defect. Can be prevented to a minimum and detection sensitivity can be increased. As a result, the defect detection sensitivity, which is the most important for the inspection device, can be increased. As a result, it has become possible to find minute defects that could not be found by the conventional method in the halftone mask. Moreover, even if the type of the halftone film changes, it can be dealt with by selecting the optimum filter each time, which is extremely effective for the inspection of the halftone mask.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of an inspection method in a pattern defect inspection apparatus.

FIG. 2 is a diagram showing a system configuration of a pattern defect inspection apparatus according to the first embodiment.

FIG. 3 is a diagram showing observation data at a pattern edge portion in a chrome mask.

FIG. 4 is a diagram showing design data and design pattern image data at a pattern edge portion.

FIG. 5 is a diagram showing observation data at a pattern edge portion in a halftone mask.

FIG. 6 is a diagram showing a transparent area for halftone calibration and a semitransparent area.

FIG. 7 is a diagram showing the principle of halftone calibration.

FIG. 8 is a diagram showing a circuit configuration for performing offset addition processing.

FIG. 9 is a diagram showing observation data at an edge portion of a halftone pattern after calibration.

FIG. 10 is a diagram showing design pattern image data at an edge portion of a halftone pattern.

FIG. 11 is a diagram showing a defect detection principle when there is a defect in a halftone pattern.

FIG. 12 is a diagram showing a sensor output when a halftone pattern has a pinhole defect.

FIG. 13 is a diagram showing observation data of pinhole defects in a halftone pattern after calibration.

FIG. 14 is a diagram showing a configuration of a data conversion circuit when an offset is added to design data.

FIG. 15 is a diagram showing a waveform obtained by applying a point spread function to design data, and a waveform obtained by further adding an offset.

FIG. 16 is a diagram showing observation data and design data at a pattern edge portion of a halftone mask.

FIG. 17 is a diagram showing an example of a blur function of a halftone mask.

FIG. 18 is a diagram showing a concept for creating inspection reference data such as a halftone mask from design data.

FIG. 19 is a block diagram showing an example of a processing process in a data conversion circuit.

FIG. 20 is a diagram showing an example of a waveform obtained by filtering observation data.

FIG. 21 is a view showing a concept for creating halftone mask field inspection reference data from design data.

[Explanation of symbols]

1 ... XYθ table 2 ... Inspected mask 3 ... Light source 4 ... Objective lens 5 ... Photodiode array 6 ... Sensor circuit 7 ... Position circuit 8 ... Comparison circuit 9 ... Magnetic disk device 10 ... Control computer 11 ... Pattern generation circuit 21 ... Data conversion circuit

Front page continuation (72) Inventor Toshiyuki Watanabe 1 Komukai Toshiba-cho, Kouki-ku, Kawasaki-shi, Kanagawa Toshiba Research & Development Center Co., Ltd. (56) Reference JP-A-6-175353 (JP, A) -150505 (JP, U) JP-B-63-3450 (JP, B1) (58) Fields investigated (Int.Cl. ⁷ , DB name) G01B 11 / 00-11 / 30

Claims (7)

(57) [Claims] 1. A sample on which a pattern is formed is irradiated with light of a predetermined wavelength, a pattern image of the sample is received by a light receiving element through an optical system, and multivalued design is made from binary design data corresponding to the pattern image. In a pattern defect inspection method for inspecting a pattern defect on a sample surface by creating pattern image data and comparing observation data corresponding to a received pattern image with design pattern image data, the observation data and the design pattern image data Before comparing with, the semi-transparent portion of the pattern on the sample surface is regarded as a non-transparent portion, and the observation data of the semi-transparent portion is made to coincide with the design pattern image data corresponding to the semi-transparent portion. and adjust the offset, and the observation data of the transparent portion of the sample surface on the pattern corresponding to the transparent portion design pattern Image As match the Jideta, after adjusting the gain of the light receiving element amplifier, the translucent portion signal than level caused by the defect of the translucent portion
Low negative signal is not cut at zero level,
While giving a predetermined offset amount to the light receiving element amplifier,
The same offset amount for the design pattern image data
A method for inspecting a pattern defect, which comprises: 2. When the sample is provided with a pattern of a semitransparent region and a transparent region which are equal to or larger than a predetermined region, and when the offset and gain of the light receiving element amplifier are adjusted, the semitransparent portion is first illuminated and then the Obtained light receiving element amplifier output (low level
Adjust the offset of the light receiving element amplifier so that the reference value) becomes zero, and then illuminate the transparent part so that the output (high level reference value) of the light receiving element amplifier obtained at that time becomes a predetermined value. 2. The pattern defect inspection method according to claim 1, wherein the gain of the light receiving element amplifier is adjusted. 3. After adjusting the offset and gain
The negative signal to the low level reference value of the light receiving element amplifier.
It is characterized by giving an offset amount larger than the amplitude of
The pattern defect inspection method according to claim 2. 4. A multivalued design pattern from binary design data
To change the conversion function when creating image data
The pattern defect inspection method according to claim 1. 5. When the optical characteristic of the pattern on the surface of the sample shows a characteristic of being semi-transparent to the irradiated light, at least three conversion functions are used to produce multivalued design pattern image data from binary design data. The pattern defect inspection method according to claim 4, wherein the function has the above peaks. 6. A means for irradiating a sample having a pattern with light of a predetermined wavelength, a light receiving element for receiving a pattern image of the sample obtained by the light irradiation by this means through an optical system, and an output of this light receiving element. A light-receiving element amplifier that amplifies light, a means for generating multivalued design pattern image data from binary design data corresponding to the pattern image of the sample, and a semitransparent portion of the pattern on the sample surface that is regarded as a nontransparent portion. Means for adjusting the offset of the light-receiving element amplifier so that the observation data of the transparent portion matches the design pattern image data corresponding to the semi-transparent portion, and the observation data of the transparent portion of the pattern on the sample surface is transferred to the transparent portion. so as to coincide with the corresponding design pattern image data, and means for adjusting the gain of the light receiving element amplifier, the offset and gain adjustment In the state, the semi-transparent
Signal level lower than semi-transparent areas caused by defect in area
The negative signal to prevent it from being cut off at zero level.
A predetermined offset amount is given to the optical element amplifier and
The same offset amount is applied to the design pattern image data.
The means for giving, the offset and the gain are adjusted , and the predetermined off
A means for inspecting a pattern defect on the sample surface by comparing the observation data obtained from the output of the light receiving element amplifier with the design pattern image data in the state where the set amount is given. A pattern defect inspection apparatus characterized by: 7. A binary setting corresponding to the pattern image of the sample.
Create multivalued design pattern image data from total data
Difference in the optical characteristics of the pattern on the sample surface
7. The conversion function according to claim 6, wherein the conversion function is changed.
On-board pattern defect inspection system.