JP2000171230A - Method and device for inspecting micropattern shape - Google Patents

Abstract

PROBLEM TO BE SOLVED: To precisely inspect a micropattern and to automatically inspect the shape of the micropattern with high precision. SOLUTION: This device is equipped with a reference image generating means 21 which generates reference image data corresponding to variance in density difference and shape difference according to pieces of good-quality image data, a difference image extracting means 22 which performs inter-image operation between the reference image data generated by the reference image generating means 12 and image data to be inspected to extract their difference image data, and a defect extracting means 23 which compares the extracted difference image data with a predetermined threshold to extract a defect part of the density and a defect part of the shape.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a micropattern shape inspection method and apparatus for measuring, for example, a defect of a micropattern formed on a semiconductor wafer, a state of a tapered portion thereof, and a dimension.

[0002]

2. Description of the Related Art Inspection of such a fine pattern formed on a semiconductor wafer is performed, for example, by using a scanning electron microscope (S
EM: Scanning Elctron Microscope) is used. This SEM uses an electron beam (electron beam)
Is used to detect secondary electrons emitted from the measurement sample at this time to obtain image data. This image data is subjected to image processing for defect inspection of the minute pattern formed on the measurement sample and this minute pattern. The state of the tapered part,
Furthermore, the dimensions are measured.

[0003] Of these, defects of the minute pattern are inspected by displaying an image of the minute pattern on a monitor in an enlarged manner and visually confirming the image displayed on the monitor by an inspector. In the defect inspection of this minute pattern, if the shape of the minute pattern has no distortion and is a perfect circle,
R-θ conversion is performed from the center of the perfect circle, and the converted image is processed to inspect the fine pattern for scratches and dirt.

[0004] The shape distortion of the minute pattern is inspected by measuring the aspect ratio, perimeter, area and the like of the minute pattern and comparing them with standard values.

In order to measure the size of a minute pattern, an image of the minute pattern is enlarged and displayed on a monitor, and an inspector puts a cursor on the image displayed on the monitor, and the distance between the cursors, the magnification, and the like. From the measurement.

[0006]

However, in the method in which the inspector visually inspects the image of the minute pattern enlarged and displayed on the monitor and inspects the defect of the minute pattern and the state of the tapered portion, the inspector evaluates the evaluation criteria. And a quantitative inspection is impossible. In addition, it is not possible to accurately inspect and evaluate a measurement sample having various shapes such as a fine pattern of a semiconductor wafer and having a large amount of deformation and distortion.

The shape distortion of the micropattern is inspected by measuring the aspect ratio, perimeter, and area of the micropattern and comparing the measured values with standard values. In this case, the shape distortion cannot be accurately expressed, and only a rough inspection can be performed.

In the method in which an inspector measures a dimension by placing a cursor on a monitor on which an image of a minute pattern is displayed, the measurement result varies depending on the position of the cursor, and high-precision measurement is impossible. is there.

Accordingly, an object of the present invention is to provide a micropattern shape inspection method and apparatus capable of inspecting a micropattern with high accuracy.

Another object of the present invention is to provide a fine pattern shape inspection method and apparatus capable of automatically and precisely inspecting the shape of a fine pattern.

Another object of the present invention is to provide a micropattern shape inspection method and apparatus capable of automatically and highly accurately inspecting the shape of a tapered portion of a micropattern having deformation and distortion in various shapes. Aim.

It is another object of the present invention to provide a micropattern shape inspection method and apparatus capable of automatically and precisely performing shape distortion of a micropattern.

Another object of the present invention is to provide a fine pattern shape inspection method and apparatus which can automatically and precisely measure the size of a fine pattern.

[0014]

According to a first aspect of the present invention, there is provided a reference image generating step of generating reference image data based on a plurality of non-defective image data according to a density difference or a variation in a shape difference. A difference image extraction step of performing an inter-image operation between the reference image data generated in the step and the image data to be inspected to extract the difference image data; and a difference image data extracted in the difference image extraction step, the difference image data being predetermined. And a defect extracting step of comparing a threshold value with a threshold value and extracting a density defect portion or a shape defect portion.

According to the second aspect, image data obtained by detecting secondary electrons emitted from the measurement sample when the measurement sample is irradiated with an electron beam is used as the image data to be inspected. In a fine pattern shape inspection apparatus for inspecting a pattern, reference image generation means for generating reference image data corresponding to a variation in density or shape difference based on a plurality of non-defective image data; A difference image extracting means for performing an inter-image operation between the reference image data and the image data to be inspected to extract the difference image data, and a difference between the difference image data extracted by the difference image extracting means and a predetermined threshold value And a defect extracting means for extracting a defect portion having a density or a defect portion having a shape by comparing the density pattern portion with the defect portion.

According to a third aspect of the present invention, in the minute pattern shape inspection apparatus according to the second aspect, the reference image generating means is formed from the maximum value for each coordinate position among the image values of the plurality of non-defective image data. Upper limit grayscale image data, lower limit grayscale image data formed from each minimum value for each coordinate position among image values of a plurality of good image data, and each coordinate position among image values of image data obtained by differentiating good image data The upper limit differential image data formed from each maximum value, the lower limit differential image data formed from each minimum value for each coordinate position among the image values of image data obtained by differentiating non-defective image data, the upper limit grayscale image data and the lower limit A function to generate average grayscale image data obtained by averaging grayscale image data and averaged differential image data obtained by averaging upper limit differential image data and lower limit differential image data To.

According to a fourth aspect of the present invention, in the minute pattern shape inspection apparatus according to the second aspect, the difference image extracting means includes a difference between the upper limit grayscale image data and the inspected image data, and a lower limit grayscale image data and the inspected image data. And the function of obtaining differential difference image data based on the difference between the upper limit differential image data and the image data to be inspected and the difference between the lower limit differential image data and the image data to be inspected. Have.

According to a fifth aspect of the present invention, in the minute pattern shape inspection apparatus according to the second aspect, the defect extracting means extracts a density defective portion by comparing the grayscale difference image data with a predetermined threshold value. And a function of comparing the differential difference image data with a predetermined threshold value to extract a shape defect portion.

According to a sixth aspect of the present invention, there is provided a minute pattern shape inspection method for inspecting a minute pattern based on image data obtained by photographing a minute pattern formed in a concave and convex shape. An inspection area setting step of setting a curved inspection area corresponding to the image data with respect to the image data, and a density peak point sequence based on a luminance distribution of a portion intersecting the bending direction of the inspection area set in the inspection area setting step. The density peak point sequence to be obtained and the density data in a range substantially perpendicular to the density peak point sequence obtained in the density peak point sequence and including the density peak points are obtained at a plurality of locations in the density peak point sequence. And obtaining predetermined values for each of the density data obtained in the obtained density data step. And the defect determination step of determining acceptability of fine patterns by comparing the quasi a fine pattern shape inspection method having.

According to the seventh aspect, the image data obtained by detecting the secondary electrons emitted from the measurement sample when the measurement sample is irradiated with the electron beam is used as the image data to be inspected. In a micropattern shape inspection apparatus for inspecting a minute pattern having a curved shape, an inspection area setting means for setting a curved inspection area corresponding to the shape of the minute pattern to image data, and an inspection area setting means. A density peak point sequence means for obtaining a density peak point sequence based on the luminance distribution of a portion crossing the curved direction of the set inspection region, and a direction substantially perpendicular to the density peak point sequence obtained by the density peak point sequence means. Density data means for obtaining density data in a range including the density peak points at a plurality of points in the density peak point sequence, and each density data obtained by the obtained density data means. And a defect determining means for determining whether or not the fine pattern is good by comparing these values with a predetermined good or bad judgment criterion. is there.

According to an eighth aspect of the present invention, in the fine pattern shape inspection apparatus according to the seventh aspect, the inspection area setting means is disposed outside and inside the edge of the minute pattern with respect to the elliptical minute pattern. It has a function of setting an inspection area formed by an ellipse.

According to the ninth aspect, in the fine pattern shape inspection apparatus according to the seventh aspect, the density peak point sequence means includes:
It has a function of setting straight lines that cross each ellipse of the inspection area at predetermined angles, obtaining each density peak position of each brightness distribution on these straight lines, and acquiring a density peak point sequence.

According to a tenth aspect of the present invention, in the minute pattern shape inspection apparatus according to the seventh aspect, the acquired density data means sets the range including the density peak point for acquiring the density data to the amount of unevenness of the fine pattern, the small amount of unevenness, The amount of pattern distortion,
It has a function to set according to the size fluctuation.

According to an eleventh aspect of the present invention, in the minute pattern shape inspection apparatus according to the seventh aspect, the defect judging means adds a luminance to each density data in a predetermined direction.
It has a function of calculating the maximum, minimum, average, and variance of the density addition value, and comparing the maximum, minimum, average, and variance values with a predetermined quality criterion to determine the quality of the fine pattern. .

According to a twelfth aspect of the present invention, there is provided a micropattern shape inspection method for inspecting a micropattern based on image data obtained by photographing a micropattern formed in an uneven and curved shape. A first shape processing step of performing a filtering process on the density peak point sequence obtained based on the luminance distribution in the inspection area according to A second shape processing step of performing a filtering process on the density peak point sequence obtained based on the density peak point sequence, and a density peak point sequence and a reference density peak point respectively filtered by the first and second shape processing steps. An evaluation step of comparing a row with a micropattern to evaluate a micropattern.

According to a thirteenth aspect of the present invention, in the fine pattern shape inspection method according to the twelfth aspect, the first and second shape processing steps include a curved inspection according to the shape of the minute pattern or the reference minute pattern. An inspection area setting step of setting an area with respect to the image data; and a density peak point sequence step of obtaining a density peak point sequence based on a luminance distribution of a portion crossing a bending direction of the inspection area set in the inspection area setting step. And a filtering process for filtering the density peak point sequence obtained in the density peak point sequence step.

According to a fourteenth aspect, in the fine pattern shape inspection method according to the thirteenth aspect, in the filtering process, the density peak point sequence is subjected to Fourier series expansion, and the shape is restored using Fourier coefficients up to a predetermined order.

According to the fifteenth aspect, the image data obtained by detecting the secondary electrons emitted from the measurement sample when the measurement sample is irradiated with the electron beam is used as the image data to be inspected. Pattern inspection apparatus for inspecting a minute pattern having a curved shape in which a density peak point sequence obtained based on a luminance distribution in an inspection area according to the shape of the minute pattern is subjected to a filtering process. Shape processing means, second shape processing means for performing a filtering process on a sequence of density peak points obtained based on the luminance distribution in the inspection area according to the shape of the reference fine pattern, and first and second processing means Evaluation means for evaluating the minute pattern by comparing the density peak point sequence filtered by the second shape processing means with the reference density peak point sequence. A small pattern shape inspection apparatus.

According to a sixteenth aspect of the present invention, in the fine pattern shape inspection apparatus according to the fifteenth aspect, the first and second shape processing means are configured to perform a curved inspection according to the shape of the minute pattern or the reference minute pattern. Inspection area setting means for setting an area with respect to image data; density peak point sequence means for obtaining a density peak point sequence based on a luminance distribution of a portion crossing a bending direction of the inspection area set by the inspection area setting means; Filter processing means for performing a filtering process on the density peak point sequence obtained by the density peak point sequence means.

According to a seventeenth aspect, in the minute pattern shape inspection apparatus according to the sixteenth aspect, the filter processing means expands the density peak point sequence into a Fourier series and restores the shape using Fourier coefficients up to a predetermined order. Has functions.

According to an eighteenth aspect of the present invention, in the micropattern shape inspection apparatus according to the fifteenth aspect, the evaluation means includes a step of forming each pattern formed by the filtered density peak point sequence and the reference density peak point sequence. It has the function of adjusting so that at least the width of the size is the same.

According to a nineteenth aspect of the present invention, in the fine pattern shape inspection apparatus according to the fifteenth aspect, the evaluation means superimposes the filtered density peak point sequence and a reference density peak point sequence, and It has a function of using the sum of the shift amounts between the patterns of the dot sequence and the reference density peak point sequence as the evaluation value of the shape distortion of the minute pattern.

According to the twentieth aspect, each edge on both sides corresponding to the contour of the minute pattern in the image signal obtained by photographing the minute pattern is obtained, and the line width of the minute pattern is determined from the interval between these edges. This is the required minute pattern shape inspection method.

According to the twenty-first aspect, a minute pattern shape inspection for inspecting a minute pattern based on an image signal obtained by detecting secondary electrons emitted from the measurement sample when the measurement sample is irradiated with an electron beam. In the apparatus, a smoothing means for smoothing a waveform with a plurality of smoothing sizes for an image signal obtained by photographing a minute pattern, and each smoothed waveform having a different smoothing size obtained by the smoothing means Differential waveform means for obtaining differential waveform data of the data; differential waveform data obtained by the differential waveform means; and each smoothed waveform data used by the differential waveform means for obtaining the differential waveform data. Width measuring means for determining each edge on both sides corresponding to the contour portion and determining the interval between these edges as the line width of the fine pattern. It is.

According to a twenty-second aspect of the present invention, in the fine pattern shape inspection apparatus according to the twenty-first aspect, a regression curve is obtained from a series of line widths of a minute pattern obtained with a smoothed size equal to or larger than a blur amount due to convolution by a diameter of an electron beam. From the value of the position where the smoothed size of the regression curve is zero, and the function of obtaining the line width of the minute pattern when the diameter of the electron beam is zero.

[0036]

(1) Hereinafter, a first embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a configuration diagram of a fine pattern shape inspection apparatus.

This fine pattern shape inspection apparatus uses an SEM
It comprises a main body 1 and a pattern inspection section 2. Among them, an electron gun 4 is provided on the electron optical column 3 of the SEM main body 1, and a condenser lens 6, a scanning coil 7 and a scanning coil 7 are provided along a traveling path of an electron beam 5 emitted from the electron gun 4. An objective lens 8 is provided. At the same time, a measurement sample 9 is provided below the electron optical column 3, and a secondary electron detector 10 is provided diagonally above the measurement sample 9.

The SEM main unit 1 is provided with a reference signal generator 11 for outputting a clock signal PS as a reference, and the clock signal PS output from the reference signal generator 11 is used for the sweep signal generator 12 and The above pattern inspection unit 2
Has been sent to

The sweep signal generator 12 is provided with the reference signal generator 1
It has a function of generating a sweep signal SS for raster-scanning the electron beam 5 based on the clock signal PS output from 1.

The magnification switching unit 13 sends a control signal C to the scanning coil 7 in combination with the setting of a magnification switching switch (not shown) and the sweep signal SS generated by the sweep signal generation unit 12.
S1 is transmitted, and the scanning coil 7 has a function of controlling the scanning direction and width of the electron beam 5.

The secondary electron detector 10 has a function of detecting secondary electrons emitted from the measurement sample 9 and outputting a detection signal of the secondary electrons. Image signal amplifier 15 connected to signal IS
Has been sent to

The image signal amplifier 15 enlarges and displays an image of a specific portion of the measurement sample 9 on the CRT display device 16 based on the image signal IS output from the amplifier 14 and the sweep signal SS generated by the sweep signal generator 12. Has a function.

On the other hand, the pattern inspection section 2 includes the CPU section 1
The A / D converter 1 is provided in the CPU 17.
8, an image signal storage unit 19 and a D / A conversion unit 20 are connected.

The A / D converter 18 includes the clock signal PS from the reference signal generator 11 and the sweep signal generator 1.
2, the image signal IS output from the amplifying unit 14 is divided into a plurality of pixels, for example, 512 × 512 pixels, by the control signal CS2 output from the CPU unit 17 based on the sweep signal SS, and the level (voltage) of the image signal IS Value) is A / D converted.

The image signal storage unit 19 includes an A / D conversion unit 18
A / D-converted image signal DIS for each pixel
And the function of recording the levels thereof.

The D / A conversion section 20 performs D / A conversion of the operation result of the CPU section 17 and performs CR conversion through the image signal amplification section 15.
It has a function of sending to the T display device 16.

The CPU section 17 reads out the image signal DIS of the designated address from the image signal recording section 19 in accordance with the inspection function main flow chart shown in FIG. 2, and performs various image processing on this image signal DIS. And a storage function.
1. It has the functions of the difference image extracting means 22 and the defect extracting means 23.

According to the reference image generation flowchart shown in FIG. 3, the reference image generation means 21 includes six pieces of reference image data corresponding to variations in density difference and shape difference based on a plurality of non-defective image data. It has a function of generating, for example, upper-limit grayscale image data, lower-limit grayscale image data, upper-limit differential image data, lower-limit differential image data, average grayscale image data, and average differential image data as non-defective pattern image data.

Specifically, as shown in FIG. 4, for example, the upper limit grayscale image data is formed from the maximum value for each coordinate position among the image values of a plurality of non-defective image data a1 to a4. The lower limit grayscale image data is formed from the minimum value for each coordinate position among the image values of a plurality of good image data, for example, as shown in FIG.

The upper limit differential image data is formed from the maximum value for each coordinate position among image values of image data obtained by differentiating non-defective image data. The lower limit differential image data is formed from the minimum value for each coordinate position among image values of image data obtained by differentiating non-defective image data.

The average grayscale image data is formed by averaging the upper limit grayscale image data and the lower limit grayscale image data.
The average differential image data is formed by averaging upper limit differential image data and lower limit differential image data.

The difference image extracting means 22 performs an inter-image operation between the six reference image data generated by the reference image generating means 21 and the image data to be inspected according to the difference image obtaining flowchart shown in FIG. Data is extracted. Specifically, based on the difference between the upper-limit grayscale image data and the image data to be inspected and the difference between the lower-limit grayscale image data and the image data to be inspected, the grayscale difference image data is obtained, and It has a function of obtaining differential difference image data based on the difference between the image data and the image data to be inspected and the difference between the lower limit differential image data and the image data to be inspected.

The defect extracting means 23 compares the difference image data extracted by the difference image extracting means 22 with a predetermined threshold value according to the defect extraction flowchart shown in FIG. The defect portion is extracted. Specifically, the density difference image data is compared with a predetermined threshold value to extract a density defect portion, and the differential difference image data and a predetermined threshold value are extracted. It has a function of comparing a value with a value to extract a defective portion of the shape.

Next, the operation of the device configured as described above will be described.

Below the electron optical column 3 of the SEM main body 1, a measurement sample 9 such as a semiconductor wafer on which a fine pattern such as an LSI is formed is placed.

When the electron beam 4 is emitted from the electron beam 5 from the electron gun 4, the electron beam 5 is reduced by the condenser lens 6 and raster-scanned in the XY directions by the scanning coil 7 in accordance with the control signal CS 1 from the magnification switching unit 13. The light is further reduced by the objective lens 8 and irradiated onto the surface of the measurement sample 9.

When the electron beam 5 is irradiated on the surface of the measurement sample 9 as described above, secondary electrons are emitted from the surface of the measurement sample 9. The secondary electron detector 10 detects the secondary electrons and outputs a detection signal. Then, this detection signal is amplified by the amplifier 14 and becomes an image signal IS, which is sent to the image signal amplifier 15.

The image signal amplifier 15 enlarges and displays an image of a specific portion of the measurement sample 9 on the CRT display 16 based on the image signal IS output from the amplifier 14 and the sweep signal SS generated by the sweep signal generator 12. .

At the same time, the image signal IS is sent to the A / D conversion unit 18 of the pattern inspection unit 2, and the A / D conversion unit 1
Reference numeral 8 denotes A / D conversion of the image signal IS based on the control signal CS2 output from the CPU unit 17, divides the image signal IS into, for example, 512 × 512 pixels by raster scanning and scanning line division, and performs A / D conversion at each address. The obtained digital image signal is represented by IS (i, j: i, j = 0, 1, 3,...)
512).

The image signal storage unit 19 includes an A / D conversion unit 18
The digital image signal (SEM image) DIS is stored for each address.

Next, the CPU section 17 controls the image signal storage section 1
The digital image signal (hereinafter referred to as image data to be inspected) DIS stored in the memory 9 is inspected for the shape of a minute pattern by using an arithmetic function. This inspection is performed by comparing the image data to be inspected DIS stored in the image signal storage unit 19 with the six reference image data previously stored in the image signal storage unit 19.

More specifically, the CPU section 17 executes step # 1 of the inspection function main flowchart shown in FIG.
In step # 2, the inspection parameters are read, and in step # 2, the sizes of the image data to be inspected and the reference image data are reset. In step # 3, preprocessing such as image reading, differentiation processing, and normalization processing is performed on the image data to be inspected. Then, the CPU section 17 reads the number of reference files in step # 4, and executes steps # 5 to # 5.
In step 8, reference image generating means 21 generates six pieces of reference image data corresponding to the density difference and the variation in shape difference.

More specifically, the reference image generation means 21 uses a plurality of reference image generation density image data and reference generation differential image data obtained by differentiating these reference image generation density image data as non-defective image data.

First, the reference image generating means 21 performs the processing shown in FIG.
In step # 20 of the reference image generation flowchart shown in FIG. 7, a matrix having, for example, a neighborhood of ± 1 as shown in FIG. 7 is used.
An image value (shade value) on the gray-scale image data for reference image generation corresponding to y) and each image value corresponding to a nearby area (fx, fy) centered on the target coordinate (x, y) are acquired.

Next, in step # 21, the reference image generation means 21 obtains the maximum image value (max) among the image values corresponding to the target coordinate (x, y) and its neighboring area (fx, fy). The maximum image value (max) and the image value (tm) at the same coordinates (x, y) of the upper limit grayscale image data area
ax), and if max> tmax, the maximum (ma) of the image value at the same coordinates (x, y) in the upper-limit grayscale image data area
x) is stored.

Next, in step # 22, the reference image generation means 21 obtains the minimum image value (min) of the image coordinates corresponding to the target coordinate (x, y) and its neighboring area (fx, fy). The minimum image value (min) is compared with the image value (tmin) of the same coordinate (x, y) of the lower limit grayscale image data area, and if min <tmin, the same coordinate of the lower limit grayscale image data area The minimum image value (min) of the image values is stored in (x, y).

Next, the reference image generation means 21 uses the matrix shown in FIG. 7 in the same manner as described above in step # 23, scans this matrix into the reference generation differential image data, and obtains the target coordinates (x, y) of the matrix. ), And the image value on the differential image data for reference generation corresponding to the coordinate of interest (x,
Each image value corresponding to the neighboring area (fx, fy) centered on y) is obtained.

Next, in step # 24, the reference image generating means 21 finds the maximum image value (max) of the image coordinates corresponding to the target coordinate (x, y) and its neighboring area (fx, fy). The maximum image value (max) is compared with the image value (tmax) of the same coordinate (x, y) in the upper limit differential image data area, and if max> tmax, the same coordinate in the upper limit differential image data area The maximum (max) of the image value is stored in (x, y).

Next, in step # 25, the reference image generation means 21 finds the minimum image value (min) of the image coordinates corresponding to the target coordinate (x, y) and its neighboring area (fx, fy). The minimum image value (min) is compared with the image value (tmin) of the same coordinate (x, y) in the lower limit differential image data area. If min <tmin, the same coordinate in the lower limit differential image data area The minimum image value (min) of the image values is stored in (x, y).

As described above, the reference image generating means 21 obtains the upper limit grayscale image data, the lower limit grayscale image data, the upper limit differential image data, and the lower limit differential image data among the six reference image data in step # 26, and obtains these images. The data is stored in the image signal storage unit 19.

Next, in step # 27, the reference image generation means 21 scans the matrix shown in FIG. 7 on the previously acquired upper-level grayscale image data in the same manner as described above, and focuses the coordinates (x, y) of this matrix. The image value (tmax) on the upper limit grayscale image data corresponding to is acquired.

Next, in step # 28, the reference image generating means 21 scans the matrix shown in FIG. 7 on the previously obtained lower-limit grayscale image data in the same manner as described above, and focuses on the coordinates (x, y) of this matrix. The image value (tmin) on the lower limit grayscale image data corresponding to is obtained.

Next, in step # 29, the reference image generation means 21 compares the image value (tmax) on the upper limit grayscale image data obtained in step # 27 with the above-mentioned step # 28.
The average value (ave) with the image value (tmin) on the lower limit grayscale image data obtained in step (1) is obtained, and the average value (ave) is defined as the average grayscale image value at the same coordinates (x,
y).

Next, in step # 30, the reference image generating means 21 scans the matrix shown in FIG. 7 on the previously acquired upper limit differential image data in the same manner as described above, and focuses on the coordinates (x, y) of this matrix. To obtain the image value (tmax) on the upper limit differential image data corresponding to.

Next, in step # 31, the reference image generation means 21 scans the matrix shown in FIG. 7 on the previously obtained lower limit differential image data in the same manner as described above, and focuses the coordinates (x, y) of the matrix. To obtain the image value (tmin) on the lower limit differential image data corresponding to.

Next, in step # 32, the reference image generating means 21 compares the image value (tmax) on the upper limit differential image data obtained in step # 30 with the above-mentioned step # 31.
The average value (ave) with the image value (tmin) on the lower limit differential image data obtained in step (1) is obtained, and the average value (ave) is used as the average differential image value, and the coordinates (x,
y).

As described above, in step # 33, the reference image generating means 21 obtains the remaining average density image data and average differential image data from the six pieces of reference image data, and stores these image data in the image signal storage unit 19. save.

Next, in step # 9 of the inspection function main flowchart shown in FIG. 2, the CPU section 17 binarizes the average differential image data among the six pieces of reference image data, as shown in FIG. And generates the reference average differential binarized image data.

Next, the CPU section 17 stores the image signal storage section 1
9 is read out, and the image data to be inspected is binarized to obtain the image data shown in FIG.
The differential binary image data to be inspected as shown in FIG.

Next, in step # 10, the CPU section 17 shifts the respective image positions of the reference average differential binarized image data and the inspected differential binarized image data shown in FIGS. 8 (a) and 8 (b). Then, an exclusive OR (XOR) of each image value of the image data is obtained, and a precise alignment for obtaining a position at which the exclusive OR (XOR) becomes minimum (minimum difference) is performed.

More specifically, the CPU 17 resets the sizes of the reference averaged differential binarized image data and the inspected differential binarized image data in step # 40 in accordance with the precision alignment flowchart shown in FIG. Next step # 41
Initialize the difference minimum counter in the next step #
At 41, a difference counter is initialized.

Next, in step # 43, the CPU 17 determines whether the reference average differential binarized image data and the differential
Exclusive OR (XOR) of the same coordinate value with the binarized image data
Ask for.

Next, in step # 44, the CPU 17 calculates the logical product (AND) of the XOR calculation result and the same coordinate value of the shielding mask image, and in the next step # 45, stores the AND result in the difference counter. to add.

Next, at step # 46, the CPU section 17 compares the count value of the difference counter with the count value of the minimum difference counter. If the count value of the difference counter is smaller than the count value of the minimum difference counter, the CPU 17 proceeds to step # 46. In step # 47, the reference image shift amount is stored, and in the next step # 48, the count value of the difference counter is set to the count value of the difference minimum counter. Thus, the reference average differential binarized image data and the inspected differential binarized image data are aligned.

Next, the difference image extracting means 22 of the CPU 17
In step # 11 of the inspection function main flowchart shown in FIG. 2, a calculation is performed between each image of the reference image data having the allowable range information of the density difference and the shape difference and the image data to be inspected to obtain The difference image containing the abnormality is extracted.

Specifically, according to the difference image acquisition flow chart shown in FIG. 5, the difference image extraction means 22 performs the same processing as described above for the dark and light inspection image data as shown in FIG. 7 is scanned, and an image value (p) corresponding to the target coordinates (x, y) of the matrix is acquired.

Next, the difference image extracting means 22 executes step #
At 51, the image values (max and max) of the coordinates (x + dx, y + dy) obtained by adding the coordinates (dx, dy) of the alignment data to the target coordinates (x, y) of the upper limit grayscale image data as shown in FIG. To get).

Next, the difference image extracting means 22 executes step #
At 52, the image value (p) of the shaded image data to be inspected
Is set as the density upper limit difference (maxtemp). That is, maxtemp = p-max (1) is calculated, and when maxtemp <0, maxtemp = 0.

Next, the difference image extracting means 22 executes step #
At 53, the matrix shown in FIG. 7 is scanned in the same manner as described above with respect to the image data to be inspected as shown in FIG. ) To get.

Next, the difference image extracting means 22 executes step #
At 54, the coordinate of interest (x,
The image value (min) of the coordinates (x + dx, y + dy) obtained by adding the coordinates (dx, dy) of the alignment data to y) is obtained.

Next, the difference image extracting means 22 executes step #
At 55, the image value (p) of the shaded image data to be inspected
And the image value (min) of the lower limit grayscale image data is set as the density lower limit difference (mintemp). That is, mintemp = min-p (2) is calculated, and when mintemp <0, mintemp = 0.

Next, the difference image extracting means 22 executes step #
At 56, the upper density difference (maxtemp) and the lower density difference (m
intemp) to obtain a density difference (d) d = maxtemp + mintemp (3) and store it in the coordinates (x, y) of the density difference image data area.

FIG. 11 is a schematic diagram showing the operation of the grayscale difference image data obtained as described above.
xtemp), a density lower limit difference (mintemp), and an image value (p) of the image data to be inspected (p).

Next, the difference image extracting means 22 executes step #
At 57, the image value (p) of the coordinate of interest (x, y) of the shaded image data to be inspected is acquired.

Next, the difference image extracting means 22 executes step #
At 58, the coordinates of interest (x,
An image value (max) of coordinates (x + dx, y + dy) obtained by adding the coordinates (dx, dy) of the alignment data to y) is obtained.

Next, the difference image extracting means 22 executes step #
At 59, the image value (p) of the shaded image data to be inspected
Is set as the differential upper limit difference (maxtemp). That is, maxtemp = p-max (4) is calculated, and when maxtemp <0, maxtemp = 0.

Next, the difference image extracting means 22 executes step #
At 60, the matrix shown in FIG. 7 is scanned with respect to the image data to be inspected in the same manner as described above, and an image value (p) corresponding to the target coordinate (x, y) of the matrix is obtained.

Next, the difference image extracting means 22 executes step #
At 61, the target coordinates (x,
The image value (min) of the coordinates (x + dx, y + dy) obtained by adding the coordinates (dx, dy) of the alignment data to y) is obtained.

Next, the difference image extracting means 22 executes step #
At 62, the image value (p) of the shaded image data to be inspected
The difference between the lower limit differential image data and the image value (min) of the lower limit differential image data is set as the differential lower limit difference (mintemp). That is, mintemp = min-p (5) is calculated, and when mintemp <0, mintemp = 0.

Next, the difference image extracting means 22 executes step #
At 63, the upper differential value (maxtemp) and the lower differential value (m
intemp) to obtain a differential difference (d) d = maxtemp + mintemp (6) and store it in the coordinates (x, y) of the differential difference image data area.

Next, the difference image extracting means 22 executes step #
At 64, the coordinates (x, y) of the grayscale difference image data area
And the image value (bd) of the coordinates (x, y) of the differential difference image data area are obtained, and these image values (nd) and image values (bd) are compared. .

As a result of this comparison, if the image value (nd) is smaller than the image value (bd) and nd <bd, the difference image extracting means 22 proceeds to step # 65 and sets nd to the coordinates of the minimum difference image area. The image data is stored in (x, y), that is, the gray level difference image value is output as the absolute value difference image value.

As a result of the comparison, if the image value (nd) is larger than the image value (bd) and nd ≧ bd, the difference image extracting means 22 proceeds to step # 66 and sets bd to the minimum difference image area. , That is, the differential difference image value is output as the absolute value difference image value.

As described above, in step # 67, the difference image extracting means 22 obtains the density difference image data, the differential difference image data, and the absolute value difference image data, and
Save to 9.

Next, the defect extracting means 23 of the CPU 17
Receives the grayscale difference image data, the differential difference image data and the absolute value difference image data extracted by the difference image extracting means 22 in step # 12 of the inspection function main flowchart shown in FIG. A defect part of the density is extracted by comparing with a predetermined threshold value,
In addition, the differential image data is compared with a predetermined threshold value to extract a defective portion of the shape.

Specifically, according to the defect extraction flowchart shown in FIG. 6, the defect extracting means 23 scans the matrix shown in FIG. x,
An image value (np) corresponding to y) is obtained.

Next, the defect extracting means 23 determines in step # 7
In step 1, the image value (np) on the grayscale image data is compared with a predetermined binarization threshold (isp.Sth: defect output threshold). The process proceeds to # 72, where it is determined that the defect is not a defect, and “0” is stored in the coordinates (x, y) of the gray-scale defect extraction image area.

The image value (np) and the binarization threshold (is
As a result of comparison with (p.Sth), if np ≧ isp.Sth, the defect extracting means 23 proceeds to step # 73, determines that the image is a defect, and changes the image value (np) to the density defect extracted image area. Store at coordinates (x, y).

Next, the defect extracting means 23 determines in step # 7
In step 4, similarly to the above, the matrix shown in FIG. 7 is scanned over the differential difference image data, and an image value (np) corresponding to the target coordinate (x, y) of the matrix is obtained.

Next, the defect extracting means 23 determines in step # 7
At 5, the image value (np) on the differential difference image data is compared with a predetermined binarization threshold (isp.Sth),
If np <isp.Sth, the flow shifts to step # 76 to judge that the defect is not a defect, and stores "0" in the coordinate (x, y) of the differential defect extraction image area.

The image value (np) and the binarization threshold (is
As a result of comparison with p.Sth), if p ≧ isp.Sth, the defect extracting means 23 proceeds to step # 77 and determines that the image is a defect, and sets the image value (np) to the differential defect extraction image area. Store at coordinates (x, y).

Next, the defect extracting means 23 determines in step # 7
At 8, the image data of density defect extraction and the image data of differential defect extraction are stored in the image signal storage unit 19.

The defect extracting means 23 scans the matrix shown in FIG. 7 on the absolute value difference image data in the same manner as described above, and obtains an image value (np) corresponding to the target coordinate (x, y) of the matrix. I do.

Next, the defect extracting means 23 compares the image value (np) on the absolute value difference image data with a predetermined binarization threshold (isp.Sth), and satisfies np <isp.Sth. if there is,
It is determined that the image is not a defect, and “0” is stored in the coordinates (x, y) of the absolute value defect extraction image area. If np ≧ isp.Sth, it is determined that the image is a defect and the image value (np) is determined. Is stored in the coordinates (x, y) of the absolute value defect extraction image area.

As described above, when the gray-scale defect extracted image data, the differential defect extracted image data, and the absolute value defect extracted image data are extracted, the CPU unit 17 determines the characteristics of the extracted regions in these image data, that is, the total area of the extracted regions, Measure the area of the maximum area, the aspect ratio of the maximum area, the total number of extraction areas,
The pass / fail of these features is determined by comparing them with a predetermined pass / fail determination criterion.

As described above, in the first embodiment, the reference image data corresponding to the variation in the density difference and the shape difference is generated based on the plurality of non-defective image data, and the reference image data and the image data to be inspected are generated. The difference image data is extracted by performing an inter-image calculation of the difference pattern data, and the difference image data is compared with a predetermined threshold value to extract a density defect portion or a shape defect portion. Shape inspection can be automatically performed with high accuracy. For example, without being affected by unevenness or fluctuation of the brightness distribution of a secondary electron image specific to a SEM image, only a shape defect of a minute pattern can be extracted and inspected, and when applied to a semiconductor process such as an LSI or a super LSI, Product evaluation and inspection can be performed easily and with high accuracy.

Note that the first embodiment may be modified as follows.

For example, in the generation of the reference image data, the vicinity area as shown in FIG. 7 is set to the area of ± 1 pixel of the pixel of interest. However, the present invention is not limited to this, and is set according to the width of the acceptable range of the fine pattern. Good.

In FIG. 8 showing the alignment for comparison between the reference image data and the image data to be inspected, the positioning movement width is set to 1 for both mx and my, but is not limited to this.
The moving amount may be set in accordance with the positional deviation amount of the images.

Although differential binary image data is used for positioning, the present invention is not limited to this. Differential image data, image data obtained by binarizing gray image data, and gray image data itself may be used for positioning.

Further, the measurement sample 9 is not limited to the LSI pattern, and can be applied to the inspection of the shape of a minute pattern for any measurement sample as long as the measurement is on the order of μm.

Further, the difference image extracting means 22 may execute a flowchart shown in FIG. That is, the difference image extracting means 22 scans the matrix shown in FIG. 7 in the same manner as described above with respect to the dark and light inspection image data as shown in FIG. An image value (p) corresponding to (x, y) is obtained.

Next, the difference image extracting means 22 executes step #
At 81, the image value (max) of the coordinates (x + dx, y + dy) obtained by adding the coordinates (dx, dy) of the alignment data to the target coordinates (x, y) of the upper limit grayscale image data as shown in FIG. To get.

Next, the difference image extracting means 22 executes step #
At 82, the image value (p) of the shaded image data to be inspected
Maxtemp, the difference between the maximum grayscale image data and the image value (max)
= P-max is set as the density upper limit difference (maxtemp).

Next, the difference image extracting means 22 executes step #
At 83, the target coordinates (x,
The image value (min) of the coordinates (x + dx, y + dy) obtained by adding the coordinates (dx, dy) of the alignment data to y) is obtained.

Next, the difference image extracting means 22 executes step #
At 84, the image value (p) of the shaded image data to be inspected
Mintemp, the difference between the lower limit grayscale image data and the image value (min)
= Min-p is set as the density lower limit difference (mintemp).

Next, the difference image extracting means 22 executes step #
At 85, the upper density difference (maxtemp) and the lower density difference (m
intemp) to obtain a density difference d = maxtemp + mintemp, and store it in the coordinates (x, y) of the density difference image data area.

Next, the difference image extracting means 22 executes step #
At 86, the image value (p) of the coordinate of interest (x, y) of the differential image data to be inspected is acquired.

Next, the difference image extracting means 22 executes step #
At 87, the target coordinates (x,
An image value (max) of coordinates (x + dx, y + dy) obtained by adding the coordinates (dx, dy) of the alignment data to y) is obtained.

Next, the difference image extracting means 22 executes step #
At 88, the image value (p) of the differential inspected image data
Maxtemp, the difference between the maximum differential image data and the image value (max)
= P-max is set as the differential upper limit difference (maxtemp).

Next, the difference image extracting means 22 executes step #
At 89, the target coordinates (x,
The image value (min) of the coordinates (x + dx, y + dy) obtained by adding the coordinates (dx, dy) of the alignment data to y) is obtained.

Next, the difference image extracting means 22 executes step #
At 90, the image value (p) of the differential inspected image data
Mintemp, the difference between the lower limit differential image data and the image value (min)
= Min-p is set as the differential lower limit difference (mintemp).

Next, the difference image extracting means 22 executes step #
At 91, the upper differential value (maxtemp) and the lower differential value (m
intemp) to obtain the differential difference d = maxtemp + mintemp, and store it at the coordinates (x, y) of the differential difference image data area.

As described above, the difference image extracting means 22 acquires the grayscale difference image data and the differential difference image data and stores them in the image signal storage section 19.

(2) Next, a second embodiment of the present invention will be described with reference to the drawings. The same parts as those in FIG. 1 are denoted by the same reference numerals, and detailed description thereof will be omitted.

FIG. 13 is a configuration diagram of a minute pattern shape inspection apparatus.

The CPU section 17 reads the image signal DIS of the designated address from the image signal recording section 19,
It has a calculation function and a storage function for performing various image processing on the image signal DIS. In particular, the inspection area setting means 30, the density peak point sequence means 31, the acquired density data means 3
2 and the respective functions of the defect determination means 33.

The inspection area setting means 30 includes a curved inspection area corresponding to the shape of the fine pattern formed on the measurement sample 9, for example, the semiconductor resist hole pattern 40 as shown in FIGS. 14 and 15. For example, it has a function of setting a taper inspection region Q as shown in FIG. 16 with respect to image data obtained by the SEM main body 1.

Here, the taper inspection region Q is composed of an ellipse 1 and an ellipse 2. The parameters of the ellipse 1 and the ellipse 2 are the ellipse 1 (the major axis A1, the minor axis B1) and the ellipse 2 (the major axis A2, The minor diameter B2) is set, and the ellipse 1 includes the resist hole pattern 40 and the ellipse 2 includes the resist hole pattern 40 in consideration of the size variation, the shape distortion, and the pattern alignment error of the reference resist hole pattern 40. It is set in advance so as to be included in the resist hole pattern 40.

The density peak point sequence means 31 is a portion crossing the direction of curvature of the taper inspection area Q set by the inspection area setting means 30, for example, as shown in FIG.
Straight lines L1 and L2 crossing each ellipse ,..., Ln are set for each predetermined angle, and these straight lines L1 , L2,..., Ln, the respective density peak positions are obtained from the respective brightness distributions, and a density peak point sequence (Xi, Yi) as shown in FIG. 17 is obtained. Note that i is 1 to M.

As shown in FIG. 18, the acquired density data means 32 is substantially perpendicular to the density peak point sequence (Xi, Yi) obtained by the density peak point sequence means 31, that is, each line Pi, Pi + 1, , And a function of acquiring density data in a range including a density peak point at the peak coordinates (Xi + 1, Yi + 1) (hereinafter referred to as a density data acquisition range) ± α as shown in FIG. Have.

The acquired density data means 32 has a function of setting the density data acquisition range ± α in accordance with the amount of unevenness of the resist hole pattern 40, the amount of distortion of the resist hole pattern 40, and the size variation thereof. I have.

The defect judging means 33 adds the luminance to each of the density data acquired by the acquired density data means 32 in a predetermined direction, and calculates the maximum, minimum, average, and variance of the density addition value. It has a function of determining the quality of the resist hole pattern 40 by comparing the maximum, minimum, average, and variance values with a predetermined quality determination criterion.

Next, the operation of the device configured as described above will be described.

In the lower part of the electron optical column 3 of the SEM main body 1, a minute resist hole pattern 40 such as an LSI is provided.
A measurement sample 9 such as a semiconductor wafer on which is formed is placed.

When the electron beam 4 is emitted from the electron beam 5 from the electron gun 4, this electron beam 5 is reduced by the condenser lens 6 and rasterized in the X-Y direction by the scanning coil 7, as in the first embodiment. It is scanned, further reduced by the objective lens 8 and irradiated onto the surface of the measurement sample 9.

The secondary electrons emitted from the surface of the measurement sample 9 are detected by the secondary electron detector 10, and the detection signal is sent to the image signal amplifier 15 as the image signal IS through the amplifier 14. The resist hole pattern 40 is enlarged and displayed on the CRT display device 16 by the image signal amplifier 15.

At the same time, the image signal IS is A / D converted by the A / D converter 18 and divided into, for example, 512 × 512 pixels by raster scanning and scanning line division.
(I, j: i, j = 0, 1, 3,... 512). The image signal storage unit 19 stores these digital image signals (SEM images) DIS for each address.

Next, the CPU section 17 stores the image signal in the image signal storage section 1.
The resist hole pattern 40 is inspected by using an arithmetic function on the digital image signal (image data to be inspected) DIS stored in 9.

FIGS. 14 and 15 are image examples of the resist hole pattern 40 obtained by the SEM main body 1, and FIG.
4 to FIG. 15, the sagging (taper) of the resist shape at the edge of the resist hole pattern 40 becomes remarkable, thereby causing a change in the image density on the SEM image.

In the SEM image shown in FIG. 14 where the taper is normal, the brightness indicating the taper at the hole edge is uniform and the width is narrow. However, in the SEM image shown in FIG. 15 in which the taper is abnormal, the uniformity of luminance is lost, the width is narrow, and the luminance itself is high.

In such a hole taper inspection of the resist hole pattern 40, the brightness of the hole edge portion of the SEM image reflecting the tapered state of the hole edge is accurately evaluated.

First, the inspection hole image data DIS stored in the image signal storage unit 19 includes the resist hole pattern 40.
Are present, a resist hole pattern 40 to be measured is selected from the resist hole patterns 40, and the position thereof is specified.

The CPU 17 uses the hole image shown in FIG. 20 as the template image or the reference image S in FIG.
Then, a normalized cross-correlation operation with the image data DIS to be inspected shown in FIG. 21 is performed, and a position where the similarity is equal to or more than a specified value is defined as a hole pattern position (Cx, Cy) as shown by “+” in FIG.

Next, the inspection area setting means 3 of the CPU 17
0 is the center of the hole pattern position (Cx, Cy) of the selected and specified resist hole pattern 40, and each ellipse 1 and 2 of the taper inspection region Q shown in FIG. Set for 40.

Next, the inspection area setting means 30 sets an area inside the ellipse 1 and outside the ellipse 2 in the taper inspection area Q set for the resist hole pattern 40 as an inspection area.

Next, in the taper inspection region Q set with respect to the resist hole pattern 40, the density peak point sequence means 31 moves from the center (C _x , C _y ) of the hole pattern 40 as shown in FIG. Each straight line L1 crossing the ellipses 1 and 2,
L2,..., Ln are set for each predetermined angle.
, L2,..., Ln are obtained. FIG.
(a) to (c) show respective luminance distributions on the straight lines L1, L2,..., Ln.

Next, the density peak point sequence means 31 finds each density peak position from each luminance distribution on each of the straight lines L1, L2,..., Ln, and obtains the tapered portion of the resist hole pattern 40 as shown in FIG. Concentration peak sequence (Xi, Yi)
To get.

Next, the acquired density data means 32 outputs the density peak point sequence (Xi,
18, a density peak point sequence (Xi, Yi) (Xi + 1, Yi + 1) (Xi +) as shown in FIG.
2, Yi + 2) (Xi + 3, Yi + 3), the gradient A connecting each density peak point (Xi, Yi) to (Xi + 1, Yi + 1).
A straight line Pi having an inclination 1 / Ai that intersects the straight line Ki of i in the vertical direction is obtained.

Similarly, each density peak point (Xi + 1, Yi + 1)
And (Xi + 2, Yi + 2), a straight line Pi + 1 having a slope 1 / Ai + 1 that intersects perpendicularly with a straight line Ki + 1 having a slope Ai + 1. Further, each of the concentration peak points (Xi + 2, Yi + 2) and (Xi +
3, Yi + 3) and a straight line Pi + 2 having a slope 1 / Ai + 2 that intersects perpendicularly with a straight line Ki + 2 having a slope Ai + 2.

Hereinafter, similarly, each concentration peak point (XM-1,
YM-1) and a straight line KM-1 of the slope AM-1 connecting (XM, YM)
A straight line with a slope of 1 / AM-1
Find -1.

Next, the acquired density data means 32 calculates each peak coordinate (Xi + 1, Yi + 1) to (XM-1, YM-1) as shown in FIG. 19 on each of the straight lines Pi to PM-1. The density data in the density data acquisition range ± α including the density peak point is obtained.

Thus, by resetting the density data acquisition range ± α from the peak coordinates of the taper luminance, the tapered portion of the resist hole pattern 40 can be formed without being affected by the size fluctuation or shape distortion of the resist hole pattern 40. Brightness information can be obtained accurately.

FIG. 24 is a schematic diagram showing luminance information of the tapered portion of the resist hole pattern 40 in the density data acquisition range ± α. The acquired density data means 32 obtains acquired density data G _i (n) (where n Develops and aligns 1 to 2 × α + 1) to generate a luminance image of the tapered portion.

Next, the defect judging means 33 adds the luminance to each of the acquired density data acquired by the acquired density data means 32 in a predetermined direction, and calculates the maximum, minimum, average, dispersion Is calculated.

Next, as shown in FIG. 25 and FIG. 26, the defect judging means 33 compares the maximum, minimum, average, and variance values with predetermined pass / fail judgment criteria to determine the resist hole pattern 4.
The quality of 0 is determined.

As described above, in the second embodiment, the taper inspection region Q corresponding to the shape of the resist hole pattern 40 is set for the image data DIS to be inspected, and each straight line traversing the taper inspection region Q is set. A density peak point sequence is obtained based on each of the above luminance distributions, density data is obtained in a density data acquisition range ± α substantially perpendicular to the density peak point sequence and including the density peak point, Since the maximum, minimum, average, variance, etc. of the density addition value are calculated and obtained for the data, these values are compared with a predetermined quality judgment criterion to judge the quality of the resist hole pattern 40. In addition, it is possible to extract only the defective portion of the tapered portion of the resist hole pattern 40 having various shapes and deformation or distortion, and to perform the inspection with high precision, quantitatively and automatically. Thus, when applied to a semiconductor process such as an LSI or a super LSI, product evaluation and inspection can be performed easily and with high accuracy.

The second embodiment may be modified as follows.

For example, the measurement sample 9 has a shape close to an ellipse like the resist hole pattern 40, but is not limited to this, and may be a convex body if the contour of the pattern shape is a closed curve. If the contour of the pattern shape is not a closed curve, it can be applied to inspection if it is a part of a convex body.

Although the density peak coordinates of the tapered portion are used only for developing the brightness in the specified range of the tapered portion, the width W, height H, and aspect ratio H of the measurement sample 9 are determined using a point sequence of the density peak coordinates.
/ W, perimeter P, area S, roundness P ² / (4πS), and the like can be obtained and used for shape evaluation of the measurement sample 9, that is, distortion evaluation.

Although an LSI pattern has been described as an example of the measurement sample 9, the present invention is not limited to this, and the present invention can be applied to the inspection of the shape of a minute pattern for any measurement sample as long as measurement is on the order of μm.

(3) Next, a third embodiment of the present invention will be described with reference to the drawings. The same parts as those in FIG. 13 are denoted by the same reference numerals, and detailed description thereof will be omitted.

FIG. 27 is a block diagram of a fine pattern shape inspection apparatus.

The CPU section 17 includes an image signal storage section 19
Is read out from the inspection target image data DIS stored in the memory, and the luminance in the inspection area corresponding to the shape of the minute pattern formed on the measurement sample 9, for example, the shape of the resist hole pattern 40 of the semiconductor, is read from the inspection target image data DIS. Filter processing is performed on the concentration peak point sequence determined based on the distribution, while the shape of the resist hole pattern 40 as a reference selected from the plurality of semiconductor resist hole patterns 40 formed on the measurement sample 9. Filter processing is performed on the density peak point sequence determined based on the luminance distribution in the inspection area according to the above, and the filtered density peak point sequence to be evaluated is compared with the reference density peak point sequence. The inspection area setting means 30, the density peak point sequence means 31, and the It has the functions of data processing means 50 and evaluation means 51.

The inspection area setting means 30, the density peak point sequence means 31, and the filter processing means 50 are configured to serve also as the first and second shape processing means. The first shape processing means is configured by applying a filtering process to the density peak point sequence obtained based on the luminance distribution in the inspection area corresponding to the shape of the resist hole pattern 40 according to the shape of the reference resist hole pattern 40. The second shape processing means is configured by performing a filtering process on the sequence of density peak points obtained based on the luminance distribution in the inspection area.

Here, the inspection area setting means 30 and the concentration peak point sequence means 31 have the same functions as those of the third embodiment, and the inspection area setting means 30 is provided with a semiconductor resist to be evaluated. It has a function of setting a curved inspection area corresponding to the hole pattern 40 or the standard resist hole pattern 40, for example, a taper inspection area Q as shown in FIG. ing.

The density peak point sequence means 31 is a section which crosses the curved direction of the taper test area Q set by the test area setting means 30, for example, each straight line L1 which crosses each ellipse of the taper test area Q as shown in FIG. , L2 ,..., Ln are set for each predetermined angle, and these straight lines L1 , L2, ..., L
It has a function of obtaining each density peak position (Xi, Yi) as shown in FIG. 17 by obtaining each density peak position from each brightness distribution on n.

The filtering means 50 outputs the density peak point sequence (Xi, Yi) to be evaluated or the reference density peak point sequence (Xi, Yi) acquired by the density peak point sequence means 31.
Yi) is subjected to a low-pass filter process, and specifically has a function of expanding the density peak point sequence (Xi, Yi) into a Fourier series and restoring the shape using Fourier coefficients up to a predetermined order. I have.

That is, the pattern shape obtained by each density peak point sequence (Xi, Yi) obtained by the density peak point sequence means 31 has a fine pattern shape due to noise and luminance change peculiar to the SEM image. Shape variation is included. Such a minute shape variation becomes a factor that hinders the evaluation of the pattern shape distortion, so it is necessary to remove and shape it by spectral processing. Therefore, the density peak point sequence (Xi, Yi) Filter processing.

Explaining the spectrum processing of the pattern shape, the shape of the closed curve figure as shown in FIG. 28 can be expressed as follows by the length t measured along the periphery of the figure.

X = fx (t) (7) y = fy (t) (8) where 0 ≦ t ≦ L, and fx (0) = fx (L), f
y (0) = fy (L), where L is the length of one round of the closed curve.

Since the above equations (7) and (8) are periodic functions whose period is the length L around the closed curve, they can be expanded to a Fourier series. That is, fx (t), fy
A set (spectrum) of Fourier coefficients is obtained by Fourier series expansion of (t). In this set of Fourier coefficients, the higher the order of the Fourier coefficient, the more information about the shape of the detail of the figure is included.

Therefore, by reconstructing a figure using coefficients up to an appropriate order after Fourier series expansion, minute shape fluctuations can be removed. The order of the coefficient of the Fourier series used is the S / N of the image data DIS to be inspected.
It is experimentally determined by the above.

Among them, the reference density peak point sequence (Xi, Yi) filtered by the filter processing means 50
Are registered in advance in the reference shape registration unit 52 according to a command from the CPU unit 17.

The evaluation means 51 performs a filtering process by the filtering processing means 50, that is, a density peak point sequence to be evaluated and a reference shape registration unit 52 in which a figure is restored by using coefficients up to an appropriate order after being subjected to Fourier series expansion. And has a function of evaluating the resist hole pattern 40 by comparing with a standard density peak point sequence read from the.

When comparing the density peak point sequence to be evaluated with the reference density peak point sequence, the evaluation means 51 forms the density peak point sequence to be evaluated and the reference density peak point sequence, respectively. It has a function of adjusting the horizontal width of each pattern size to be the same, and adjusting the aspect ratio of each pattern size to the same when accuracy is required.

When comparing the density peak point sequence to be evaluated with the reference density peak point sequence, the evaluation means 51 overlaps the density peak point sequence to be evaluated with the reference density peak point sequence. In addition, the shift amount d between each pattern between the density peak point sequence and the reference density peak point sequence
Is used as the evaluation value of the shape distortion of the resist hole pattern 40.

In the evaluation function g of the shift amount d, a parameter ± α for setting a dead zone such that −α <d <+ α → g (d) = 0 is set in order to ignore a minute shift. In order to emphasize and evaluate a large deviation, an operation such as g (d) = g * g or g (d) = exp (g) is executed.

The evaluation means 51 has a function of comparing the evaluation value of the shape distortion of the resist hole pattern 40 with a preset reference value for judging the acceptability, and judging the acceptability of the resist hole pattern 40.

Next, the operation of the device configured as described above will be described.

In the lower part of the electron optical column 3 of the SEM main body 1, for example, a minute resist hole pattern 40 such as an LSI is provided.
A measurement sample 9 such as a semiconductor wafer on which is formed is placed.

When emitted from the electron beam 5 from the electron gun 4, the electron beam 5 is reduced by the condenser lens 6 and rasterized in the XY directions by the scanning coil 7, as in the first embodiment. It is scanned, further reduced by the objective lens 8 and irradiated onto the surface of the measurement sample 9.

The secondary electrons emitted from the surface of the measurement sample 9 are detected by a secondary electron detector 10 and the detection signal is sent to an image signal amplifier 15 as an image signal IS through an amplifier 14. The resist hole pattern 40 is enlarged and displayed on the CRT display device 16 by the image signal amplifier 15.

At the same time, the image signal IS is A / D-converted by the A / D converter 18 and divided into, for example, 512 × 512 pixels by raster scanning and scanning line division.
(I, j: i, j = 0, 1, 3,... 512). The image signal storage unit 19 stores these digital image signals (SEM images) DIS for each address.

Next, the CPU section 17 stores the image signal
The distortion of the pattern shape of the resist hole pattern 40 is evaluated by using an arithmetic function on the digital image signal (image data to be inspected) DIS stored in 9.

FIGS. 14 and 15 are image examples of the resist hole pattern 40 obtained in the SEM main body 1, and FIG.
When the oval resist hole pattern 40 shown in FIG. 4 is selected as the reference shape, the resist hole pattern 40 shown in FIG. 15 has a noticeable distortion in the resist shape at the edge of the hole.

The resist hole pattern 4 shown in FIG.
Evaluation of the distortion of the resist shape of 0 will be described.

First, the inspection hole image data DIS stored in the image signal storage unit 19 includes the resist hole pattern 40.
Are present, a resist hole pattern 40 as a reference is selected from these resist hole patterns 40, and the position thereof is specified.

The CPU section 17 performs a normalized cross-correlation operation with the inspection image data DIS shown in FIG. 14 or FIG. 15 using the hole image shown in FIG. The above positions are defined as hole pattern positions (Cx, Cy) as indicated by “+” in FIG.

At this time, a plurality of resist hole patterns 40
The inclination of the image is checked from the relative position, and the inclination is corrected if necessary.

Next, the inspection area setting means 3 of the CPU 17
0 is the center of the hole pattern position (Cx, Cy) of the reference resist hole pattern 40 that is selected and specified.
22 are set with respect to the resist hole pattern 40 as shown in FIG. 22. In the taper inspection region Q, a region inside the ellipse 1 and outside the ellipse 2 is set. Set as an inspection area.

Next, in the taper inspection region Q set with respect to the reference resist hole pattern 40, the density peak point sequence means 31, for example, as shown in FIG.
.., Ln that cross the ellipses 1 and 2 from the center (Cx, Cy) of 0 at predetermined angles, and as shown in FIGS. 23 (a) to 23 (c), the respective straight lines L1, L2,. Straight lines L1, L2, ...,
Each luminance distribution on Ln is obtained.

Next, the density peak point sequence means 31 obtains each density peak position from each luminance distribution on each of the straight lines L1, L2,..., Ln, and uses the resist hole pattern as a reference as shown in FIG. The concentration peak point sequence (Xi, Yi) of the 40 tapered portions is obtained.

Next, the filter processing means 50 performs low-pass filtering on the reference density peak point sequence (Xi, Yi) acquired by the density peak point sequence means 31, that is, the density peak point sequence (Xi, Yi). Is subjected to Fourier series expansion, and the shape is restored using Fourier coefficients up to a predetermined order. The order of the Fourier series coefficient used at this time is experimentally determined by the S / N of the image data DIS to be inspected.

FIGS. 29 (a), (b) and (c) show the original shape and the higher order spectrum (1) of the reference concentration peak sequence (Xi, Yi), respectively.
It is a schematic diagram which shows shape restoration by a 0th order) and shape restoration by a low order spectrum (up to the 2nd order). As can be seen from these figures, in the shape restoration using the higher-order spectrum, a shape close to the original shape of the concentration peak point sequence (Xi, Yi) is restored. In the shape restoration using the lower-order spectrum, the concentration peak point sequence (Xi, Yi) is used. Is restored to a shape close to an ellipse deviating from the original shape of. Because of such shape restoration, the order of the Fourier series coefficient is determined by the S / N of the image data DIS to be inspected.
It is experimentally determined by the above.

As described above, the concentration peak point sequence (Xi, Yi)
Is subjected to Fourier series expansion, and the shape is restored using Fourier coefficients up to a predetermined order, so that the noise is peculiar to the SEM image, and the minute shape fluctuation included due to the influence of the luminance change is removed to be shaped. .

As described above, the reference density peak point sequence (X
i, Yi) are registered in advance in the reference shape registration unit 52 according to a command from the CPU unit 17.

On the other hand, the resist hole pattern 4 to be evaluated from the plurality of resist hole patterns 40 existing in the image data DIS to be inspected stored in the image signal storage unit 19.
0 is selected and its position is specified.

The CPU section 17 performs a normalized cross-correlation calculation with the inspection image data DIS shown in FIG. 14 or FIG. 15 using the hole image shown in FIG. The above positions are defined as hole pattern positions (Cx, Cy) as indicated by “+” in FIG.

At this time, a plurality of resist hole patterns 40
The inclination of the image is checked from the relative position, and the inclination is corrected if necessary.

Next, the inspection area setting means 3 of the CPU 17
0 indicates the ellipses 1 and 2 in the taper inspection region Q shown in FIG. 16 around the hole pattern position (Cx, Cy) of the resist hole pattern 40 to be evaluated which has been selected and specified as shown in FIG. In the taper inspection region Q, a region inside the ellipse 1 and outside the ellipse 2 is set as the inspection region.

Next, in the taper inspection region Q set for the resist hole pattern 40 to be evaluated, the density peak point sequence means 31 for example, as shown in FIG. 16, the center (Cx, Cy) of the hole pattern 40 , Ln crossing over each of the ellipses 1 and 2 are set at predetermined angles, and as shown in FIGS. 23 (a) to 23 (c), the straight lines L1, L2,.
Each luminance distribution on L2,..., Ln is obtained.

Next, the density peak point sequence means 31 obtains each density peak position from each luminance distribution on each of the straight lines L1, L2,..., Ln, and uses the resist hole pattern as a reference as shown in FIG. The concentration peak point sequence (Xi, Yi) of the 40 tapered portions is obtained.

Next, the filter processing means 50 performs low-pass filtering on the density peak point sequence (Xi, Yi) to be evaluated acquired by the density peak point sequence means 31, that is, the density peak point sequence (Xi, Yi). Yi) is subjected to Fourier series expansion, and the shape is restored using Fourier coefficients up to a predetermined order.

FIGS. 30 (a), 30 (b) and 30 (c) show the original shape of the density peak point sequence (Xi, Yi) to be evaluated, the shape reconstruction using higher order spectra (up to the tenth order), and the lower order spectra (2 It is a schematic diagram which shows shape restoration by the following). As can be seen from these figures, the shape restoration using the higher-order spectrum is restored to a shape close to the original shape of the density peak point sequence (Xi, Yi), whereas the shape restoration using the lower-order spectrum is performed using the density peak point sequence (Xi, Yi). The original shape of Yi) is restored and restored to a shape close to an ellipse.

Next, the evaluation means 51 receives the density peak point sequence to be evaluated in which the figure is reconstructed using coefficients up to an appropriate order after the Fourier series expansion by the filter processing means 50, Registration unit 52
The reference concentration peak point sequence to be evaluated is read out from the reference concentration peak point sequence registered in the above, and the concentration peak point sequence to be evaluated and the reference concentration peak point sequence are compared. Adjustment is performed so that the aspect ratio of each pattern size formed by the dot sequence is the same.

The evaluation means 51 is provided in the reference shape registration unit 5
The reference density peak point sequence restored by the order of the coefficient of the optimal Fourier series, for example, the 10th order, is read out from the plurality of reference density peak point sequences registered in 2.

Next, the evaluation means 51 joins the density peak point sequence to be evaluated with the reference density peak point sequence,
The sum of the shift amounts d between the patterns of the density peak point sequence and the reference density peak point sequence is determined by the resist hole pattern 4.
The shape distortion evaluation value is 0.

At this time, in the evaluation function g of the shift amount d as described above, for example, -α <d
G (d) = g * g or g (d) = exp in order to set a parameter ± α for setting a dead zone such that <+ α → g (d) = 0 or to emphasize and evaluate a large deviation. An operation such as (g) is being executed.

Next, the evaluation means 51 compares the evaluation value of the shape distortion of the resist hole pattern 40 with a preset reference value for judging the acceptability, and judges the acceptability of the resist hole pattern 40.

As described above, according to the third embodiment, the density peak point sequence obtained based on the luminance distribution in the taper inspection region Q corresponding to the shape of the resist hole pattern 40 is subjected to the low-pass filter processing, that is, the Fourier transform. After the series is expanded, the figure is restored using the coefficients up to the appropriate order, and the Fourier series is similarly obtained for the density peak point sequence obtained based on the luminance distribution in the inspection area according to the shape of the reference minute pattern. After the development, the figure is reconstructed using coefficients up to an appropriate order, and the low-pass filtered density peak point sequence to be evaluated is compared with the reference density peak point sequence to evaluate the resist hole pattern 40. Is performed, the distortion of the shape of the resist hole pattern 40 can be quantitatively detected as a shape distortion evaluation value. It is possible to perform quality determination of the resist hole pattern 40 by comparing with a preset quality decision reference value Joyugami evaluation value.

In this case, the order of the coefficients of the Fourier series at the time of restoring a figure using coefficients up to an appropriate order after low-pass filter processing, that is, Fourier series expansion for the density peak point sequence, is based on the image data DIS to be inspected. S /
N, and so on, each concentration peak point sequence (Xi,
The pattern shape obtained by Yi) includes fine shape fluctuations in the pattern shape due to the influence of noise and luminance change peculiar to the SEM image. It is possible to evaluate the pattern shape distortion with high accuracy by eliminating a factor that hinders the evaluation.

Therefore, when the present invention is applied to inspection of semiconductor processes such as LSI and VLSI, evaluation and inspection of semiconductor elements such as LSI and VLSI can be performed easily and with high accuracy.

Note that the third embodiment may be modified as follows.

For example, the measurement sample 9 has a shape close to an ellipse like the resist hole pattern 40, but is not limited to this, and may be a quasi-convex if the contour of the pattern shape is a closed curve. If the contour of the pattern shape is not a closed curve, it can be applied to inspection if it is a part of a convex body.

Although an LSI pattern has been described as an example of the measurement sample 9, the present invention is not limited to this, and the present invention can be applied to the inspection of the shape of a minute pattern for any measurement sample as long as the measurement is on the order of μm.

(4) Next, a fourth embodiment of the present invention will be described with reference to the drawings. The same parts as those in FIG. 1 are denoted by the same reference numerals, and detailed description thereof will be omitted.

FIG. 31 is a block diagram of a fine pattern shape inspection apparatus.

The CPU section 17 reads the image signal DIS of the designated address from the image signal recording section 19,
The image signal DIS has an arithmetic function for performing various image processing on the image signal DIS and a storage function.
It has the functions of the difference waveform means 61, the first width measuring means 62, and the second width measuring means 63.

The smoothing means 60 performs a function of smoothing a waveform with a plurality of smoothing sizes on an image signal IS obtained by photographing a minute pattern formed on an LSI or the like by the SEM main unit 1. have.

The difference waveform means 61 has a function of obtaining difference waveform data of each smoothed waveform data having a different smoothing size obtained by the smoothing means 60.

[0233] The first width measuring means 62 is provided by the differential waveform means 6
1. Based on the difference waveform data obtained in step 1 and the respective smoothed waveform data used in the difference waveform means 61 for obtaining the difference waveform data, the respective edges on both sides corresponding to the contour portion of the minute pattern are obtained. It has the function of determining the distance between the edges as the line width of the minute pattern.

The second width measuring means 63 obtains a regression curve from a series of line widths of minute patterns obtained with a smoothed size equal to or larger than the blur amount due to convolution with the diameter of the electron beam 5,
It has a function of calculating the line width of the minute pattern when the diameter of the electron beam 5 is zero from the value of the position where the smoothed size of the regression curve is zero.

Next, the operation of the device configured as described above will be described.

Below the electron optical column 3 of the SEM main body 1, a measurement sample 9 such as a semiconductor wafer on which a fine pattern such as an LSI is formed is placed.

When the electron beam 4 is emitted from the electron beam 5 from the electron gun 4, the electron beam 5 is reduced by the condenser lens 6 and rasterized in the X-Y direction by the scanning coil 7, as in the first embodiment. It is scanned, further reduced by the objective lens 8 and irradiated onto the surface of the measurement sample 9.

The secondary electrons emitted from the surface of the measurement sample 9 are detected by the secondary electron detector 10 and the detection signal is sent to the image signal amplifier 15 as the image signal IS through the amplifier 14. The minute pattern is enlarged and displayed on the CRT display device 16 by the image signal amplifier 15.

At the same time, the image signal IS is A / D-converted by the A / D converter 18 and divided into, for example, 512 × 512 pixels by raster scanning and scanning line division.
(I, j: i, j = 0, 1, 3,... 512). The image signal storage unit 19 stores these digital image signals (SEM images) DIS for each address.

Next, the CPU section 17 stores the image signal storage section 1
The digital image signal (image data to be inspected) DIS stored in the memory 9 is subjected to an arithmetic function to measure the size of a minute pattern.

First, the CPU section 17 stores the image signal storage section 1
9 is read out.

Next, the smoothing means 60 photographs the minute pattern formed on the LSI or the like by the SEM main unit 1 based on the image data DIS to be inspected as shown in FIG. The waveform of the one-line image signal IS obtained is smoothed with a plurality of smoothing sizes.

Specifically, an image signal of one line is converted into a waveform x (i)
(Here, i = 0,..., 511). Smoothing means 60
Performs smoothing of an image signal of one line with a plurality of smoothing sizes for a waveform x (i) and, for example, ten types of smoothed waveforms p1 (i) to p10 corresponding to these smoothed sizes.
(i) is obtained.

Smoothing waveform p1 (i) Smoothing size n = 3 Smoothing waveform p2 (i) Smoothing size n = 7 Smoothing waveform p3 (i) Smoothing size n = 11 Smoothing waveform p4 (i) Smoothing Smoothed size n = 15 smoothed waveform p5 (i) smoothed size n = 19 smoothed waveform p6 (i) smoothed size n = 23 smoothed waveform p7 (i) smoothed size n = 27 smoothed waveform p8 (i ) Smoothed size n = 31 Smoothed waveform p9 (i) Smoothed size n = 35 Smoothed waveform p10 (i) Smoothed size n = 39 where i = (n−1) / 2,. (N-1)
/ 2, When i <(N−1) / 2 or i> 511− (n−1) / 2, P _k (i) = 0.

FIGS. 33 (a) to 33 (e) show the smoothed waveforms p1 (i) to p5 (i) among the ten types of smoothed waveforms p1 (i) to p10 (i).
Is shown.

Next, the difference waveform means 61 is
0, two waveforms having different smoothed sizes are selected, and a difference between the two waveforms is obtained, that is, a waveform having a small smoothed size−a waveform having a large smoothed size is obtained.
Get.

Difference waveform S13 (i) p1 (i) −p3 (i) Difference waveform S24 (i) p2 (i) −p4 (i) Difference waveform S35 (i) p3 (i) −p5 (i) Difference waveform S46 (i) p4 (i) -p6 (i) Difference waveform S57 (i) p5 (i) -p7 (i) Difference waveform S68 (i) p6 (i) -p8 (i) Difference waveform S79 (i) p7 (i) -p9 (i) Difference waveform S81 (i) p8 (i) -p10 (i) Here, if the smoothed size k> l, i = (k-1)
/ 2, ..., when the 511- (k-1) / 2 , S lk (i) = p l (i) -p k (i) ... (1 0) i <(k-1) / 2 or i S _{lk (i)} = 0 when> 511− (k−1) / 2.

FIGS. 34 (a) to (c) show the difference waveform data S13, S24,
S35 is shown.

FIG. 35 shows each of the smoothed waveforms p1 (i) to p10.
The relationship between (i) and each of the differential waveform data S13 to S81 is shown.

Next, the first width measuring means 62 calculates the difference waveform data S13 to S81 obtained by the difference waveform means 61.
And each of the smoothed waveform data p1 (i) to p10 used by the differential waveform means 61 to obtain these differential waveform data S13 to S81.
Based on (i), each edge on both sides corresponding to the contour of the micropattern is obtained, and the interval between these edges is obtained as the line width of the micropattern.

More specifically, small-sized smoothed waveform data pl (i) and large-sized smoothed waveform data p _k (i) shown in FIG.
To explain using the difference waveform data S _lk (i), the left peak value Lmax of the large-sized smoothed waveform data p _k (i)
(Position Pmax) and the minimum value Lmin (position Pmin).

Next, a search is made for a position Ph that takes a value of (Lmax−Lmin) / 2 (11) from the left peak value Lmax of the smoothed waveform data p _k (i) to the left in the figure.

A = Pmax−α (13) is obtained from the half width | Pmax−Pmin | as α = half width × 2 (12), and B = Pmax−β (β = α × 25%) 14) Ask for.

The position C at which the minimum value between A and B of the difference waveform data S _lk (i) obtained in this way is obtained. Then, a curve is approximated between AC and a left peak position Peak L is obtained. The left peak position PeakL corresponds to the minute pattern of the small-sized smoothed waveform data pl (i) and the base boundary position (the left edge of the minute pattern).

In the same manner as above, the right peak position PeakR
(Right edge of the micropattern).

Thus, the first width measuring means 62 uses the left peak position Peak L and the right peak position Peak R to determine the size (line width) of the fine pattern base width of the smoothed waveform data _pl (i). W _lk W _lk = | Peak L−Peak R | (15) is obtained.

On the other hand, measurement of the dimension of the base width of the micropattern at the time of the beam diameter of the electron beam 5 from the secondary electron image in the defocused state will be described.

The second width measuring means 63 determines the size of the fine pattern base width corresponding to each smoothed size, for example, W13, W24, W35, W46, W57 for all the difference waveform data S13 to S81 in the same manner as described above. , W68, W79 and W81 are measured.

As shown in FIG. 37, when the fine pattern base width _Wlk is obtained for the waveform at the time of the sharp focus, the fine pattern base width _Wlk increases as the smoothed size increases.

However, in this case, since the beam diameter of the electron beam 5 cannot be physically set to zero, even if the smoothed size is zero, the blur of the beam diameter will be formed on the image by convolution with the beam diameter of the minute pattern. 37 (physical resolution limit), and as shown by the original waveform line width (beam diameter ≠ 0) “△” at the time of just focus in FIG. 37, the value becomes larger than the original pattern width. A similar phenomenon appears more remarkably during defocus. The larger the defocus, the larger the beam diameter folded into the pattern,
Line width at the time of defocus on FIG. 37 (large beam diameter)
The relationship between the pattern line width and the waveform smoothing size as shown by “□”.

Therefore, the pattern width at the beam diameter of 0 is obtained from the series of the minute pattern base width _Wlk . A regression curve is obtained from a series of pattern widths obtained with a smoothed size equal to or larger than the blur amount due to convolution by the beam diameter. In FIG. 37, a regression curve is obtained from the minute pattern base width dimensions W68, W79, and W81. Fig. 3
The pattern width “○” at the beam diameter of 0 shown in FIG. 7 can be obtained. The order of the regression curve is determined experimentally.
In this embodiment, it is primary.

As a result, the pattern width can be measured from the defocused or just-focused image with an accuracy higher than the physical resolution limit when the beam diameter is 0.

As described above, in the fourth embodiment, the respective edges on both sides corresponding to the contour of the minute pattern in the image signal obtained by photographing the minute pattern are obtained, and the distance between these edges is determined. Since the line width of the minute pattern is obtained, the dimension measurement of the minute pattern can be automatically performed with high accuracy only from the stable concentration distribution of the tapered portion of the minute pattern regardless of the unstable density portion of the underlying portion. it can. Repetition reproducibility of dimension measurement can be improved only by obtaining a fine pattern base width from only a difference waveform having a small smoothed size, for example, only difference waveform data S13.

Further, a regression curve is obtained from a series of line widths of the minute patterns obtained with a smoothed size equal to or larger than the blur amount due to the convolution by the diameter of the electron beam bundle 5, and the value of the position where the smoothed size of the regression curve is zero is obtained. , The line width of the minute pattern when the diameter of the electron beam 5 is zero is obtained, so that the pattern width can be measured from the defocused or just-focused image with an accuracy equal to or greater than the physical resolution limit at the beam diameter of 0.

Therefore, when the present invention is applied to a semiconductor process such as an LSI and a super LSI, it is possible to easily and accurately evaluate and inspect products.

Note that the fourth embodiment may be modified as follows.

For example, the smoothing of the waveform data is performed according to the above equation (7).
However, the present invention is not limited to this, and smoothing corresponding to the electron density shape of the electron beam 5 may be performed. For example, if the electron density shape of the electron beam 5 follows a normal distribution, it is appropriate to use a Gaussian filter. Similarly, the smoothing may not be performed in the time domain as in the fourth embodiment, but may be performed in the frequency domain after the Fourier transform, and the inverse Fourier transform may be performed.

In the fourth embodiment, the number of smoothed waveforms, the smoothed size, the number of difference waveforms, and the difference interval are set to specific numbers. However, the present invention is not limited to this. By increasing the number of smoothed waveforms and the number of difference waveforms, an estimated value of the pattern width can be obtained with higher accuracy. The smoothing size and the difference interval are also equal in the third embodiment, but need not be equal if the intervals are known.

If attention is paid only to the estimation of the pattern width when the beam diameter is 0, another method may be used for the part for obtaining the pattern width from each smoothed waveform. For example, without using a differential waveform, a pattern taper portion of a smoothed waveform,
The pattern width may be obtained from the intersection (left and right) of each base portion by linear approximation.

In the fourth embodiment, the dimension measurement of the pattern width in the horizontal direction has been described. However, the scanning direction of the electron beam bundle 5 is also changed for the measurement of the pattern width in the vertical and oblique directions. That can be dealt with. It can also be dealt with by changing the acquisition direction of one-line waveform data without changing the scanning direction of the electron beam 5.

In the fourth embodiment, the pattern base width is obtained. However, the present invention is not limited to this. If the waveform data obtained includes a base portion (flat portion) and a tapered portion, the pattern pitch and the pattern The top width can be determined.

Further, although one line of data is used as the waveform data, arbitrary integrated data of a plurality of lines may be used.

Although an LSI pattern has been described as an example of the measurement sample 9, the present invention is not limited to this and can be applied to the inspection of the shape of a minute pattern for any measurement sample as long as the measurement is on the order of μm.

[0274]

As described above in detail, according to the present invention, it is possible to provide a fine pattern shape inspection method and apparatus capable of inspecting a fine pattern with high accuracy.

Further, according to the present invention, it is possible to provide a fine pattern shape inspection method and apparatus capable of automatically and precisely performing a fine pattern shape inspection.

Further, according to the present invention, it is possible to provide a micropattern shape inspection method and apparatus capable of automatically and highly accurately inspecting the shape of a tapered portion of a micropattern having deformation and distortion in various shapes. .

Further, according to the present invention, it is possible to provide a minute pattern shape inspection method and apparatus capable of automatically and precisely performing a minute pattern shape distortion.

Further, according to the present invention, it is possible to provide a minute pattern shape inspection method and apparatus capable of automatically and precisely measuring the dimension of a minute pattern.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram showing a first embodiment of a micropattern shape inspection apparatus according to the present invention.

FIG. 2 is a main flowchart of an inspection function in the apparatus.

FIG. 3 is a reference image generation flowchart in the apparatus.

FIG. 4 is a schematic diagram showing upper-limit grayscale image data and lower-limit grayscale image data among the reference image data created by the reference image generation means in the same device.

FIG. 5 is a flowchart for acquiring a difference image in the apparatus.

FIG. 6 is a flowchart of defect extraction in the apparatus.

FIG. 7 is a schematic diagram of a matrix used in a reference image generation unit in the apparatus.

FIG. 8 is a schematic diagram of reference average differential binarized image data and inspected differential binarized image data generated by the apparatus.

FIG. 9 is a flow chart of precision alignment in the apparatus.

FIG. 10 is a schematic diagram of image data to be inspected and upper limit grayscale image data used in the apparatus.

FIG. 11 is a schematic diagram showing the operation of the grayscale difference image data obtained by the apparatus.

FIG. 12 is a flowchart of another example of acquiring a difference image in the apparatus.

FIG. 13 is an overall configuration diagram showing a second embodiment of a micropattern shape inspection apparatus according to the present invention.

FIG. 14 is a view showing an image of a semiconductor hole pattern obtained by the SEM main body of the apparatus.

FIG. 15 is a view showing an image of a semiconductor hole pattern obtained by the SEM body of the apparatus.

FIG. 16 is a schematic diagram showing a taper inspection region set on a semiconductor hole pattern image by the same device.

FIG. 17 is a schematic diagram showing a sequence of density peak points obtained from each luminance distribution on each straight line crossing the taper inspection region by the same device.

FIG. 18 is a diagram showing an operation of acquiring density data by an acquired density data unit in the apparatus.

FIG. 19 is a schematic diagram showing a density data acquisition range by an acquisition density data unit in the same device.

FIG. 20 is a schematic diagram showing a reference image for selecting and specifying a resist hole pattern 40 in the apparatus.

FIG. 21 is a schematic diagram showing a hole pattern position specified in a reference image by the device.

FIG. 22 is a schematic diagram in which each ellipse in a taper inspection region is set for a resist hole pattern in the same apparatus.

FIG. 23 is a schematic diagram showing each luminance distribution on each straight line crossing a taper inspection region set in a resist hole pattern by the same device.

FIG. 24 is a schematic diagram showing luminance information of a tapered portion of a resist hole pattern in a density data acquisition range in the same device.

FIG. 25 is a diagram showing a determination operation of a defect determination unit in the apparatus.

FIG. 26 is a diagram showing a determination operation of a defect determination unit in the apparatus.

FIG. 27 is an overall configuration diagram showing a third embodiment of the micropattern shape inspection apparatus according to the present invention.

FIG. 28 is a view for explaining the spectrum processing of the pattern shape in the same device.

FIG. 29 is a schematic diagram showing an original shape of a reference concentration peak point sequence obtained by the same device, shape restoration by a higher-order spectrum, and shape restoration by a lower-order spectrum.

FIG. 30 is a schematic diagram showing an original shape of a concentration peak point sequence to be evaluated, a shape restoration using a higher-order spectrum, and a shape restoration using a low-order spectrum, which are obtained by the apparatus.

FIG. 31 is an overall configuration diagram showing a fourth embodiment of a micropattern shape inspection apparatus according to the present invention.

FIG. 32 is a photograph 1 taken by the SEM main unit 1 of the apparatus.
The figure which shows the image signal of a line.

FIG. 33 is a schematic diagram of a smoothed waveform obtained by a smoothing means of the apparatus.

FIG. 34 is a schematic diagram of a differential waveform obtained by differential waveform means of the same device.

FIG. 35 is a view showing the relationship between each smoothed waveform obtained by the same device and each difference waveform data.

FIG. 36 is a schematic view for explaining an operation of measuring a line width of a minute pattern by the apparatus.

FIG. 37 is a schematic diagram for explaining the operation of measuring the micropattern base width when the beam diameter of the electron beam is 0 from the secondary electron image in a defocused state by the same device.

[Explanation of symbols]

 1: SEM main unit, 2: pattern inspection unit, 3: electron optical column, 4: electron gun, 6: condenser lens, 7: scanning coil, 8: objective lens, 9: measurement sample, 10: secondary electron detection 11: Reference signal generator, 12: Sweep signal generator, 15: Image signal amplifier, 16: CRT display device, 17: CPU, 18: A / D converter, 19: Image signal storage, 20: D / A conversion unit, 21: reference image generation unit, 22: difference image extraction unit, 23: defect extraction unit, 30: inspection area setting unit, 31: density peak point sequence unit, 32: acquired density data unit, 33: Defect determination means, 40: resist hole pattern, 50: filter processing means, 51: evaluation means, 52: reference shape registration unit, 60: smoothing means, 61: difference waveform means, 62: first width measurement means, 63 : Second Width measuring means.

Claims (22)

[Claims] A reference image generating step of generating reference image data corresponding to a variation in a density difference or a shape difference based on a plurality of non-defective image data; A difference image extraction step of performing an inter-image operation with the inspection image data to extract the difference image data, and comparing the difference image data extracted in the difference image extraction step with a predetermined threshold value A defect extracting step of extracting a defective portion having a density or a defective portion having a shape. 2. Inspection of a minute pattern formed on the measurement sample using image data obtained by detecting secondary electrons emitted from the measurement sample when the measurement sample is irradiated with an electron beam as image data to be inspected. A fine pattern shape inspection apparatus for performing a reference image generating means for generating reference image data according to a density difference or a variation of a shape difference based on a plurality of non-defective image data; and the reference generated by the reference image generating means. Difference image extracting means for performing an inter-image operation between image data and the image data to be inspected to extract the difference image data; and a predetermined threshold value with the difference image data extracted by the difference image extracting means. And a defect extracting means for extracting a defective portion having a density or a defective portion having a shape by comparing with the above. 3. The reference image generating means includes: upper limit grayscale image data formed from the maximum value for each coordinate position among the image values of the plurality of non-defective image data; Among the lower limit gradation image data formed from each minimum value for each coordinate position, the upper limit differential image data formed from each maximum value for each coordinate position among image values of image data obtained by differentiating the non-defective image data, The lower limit differential image data formed from each minimum value for each coordinate position among the image values of the image data obtained by differentiating the non-defective image data, the average density obtained by averaging the upper density image data and the lower density image data Image data; average differential image data obtained by averaging the upper limit differential image data and the lower limit differential image data. Item 3. A micropattern shape inspection apparatus according to Item 2. 4. The difference image extracting means, based on a difference between the upper limit grayscale image data and the image data to be inspected, and a difference between the lower limit grayscale image data and the image data to be inspected, And a function of obtaining differential difference image data based on a difference between the upper limit differential image data and the image data to be inspected, and a difference between the lower limit differential image data and the image data to be inspected. The micropattern shape inspection apparatus according to claim 2. 5. The defect extracting means extracts a density defect portion by comparing the density difference image data with a predetermined threshold value, and extracts the density difference portion from the differential difference image data. 3. The micropattern shape inspection apparatus according to claim 2, further comprising a function of comparing a value with a value to extract a defective portion of the shape. 6. A minute pattern shape inspection method for inspecting the minute pattern based on image data obtained by photographing a minute pattern formed in an uneven and curved shape, the method comprising: An inspection area setting step of setting a curved inspection area with respect to the image data; and obtaining a density peak point sequence based on a luminance distribution of a portion of the inspection area crossing a bending direction set in the inspection area setting step. A density peak point sequence step; and density data in a direction substantially perpendicular to the concentration peak point sequence obtained in the concentration peak point sequence step and in a range including the concentration peak point, at a plurality of points in the concentration peak point sequence. And obtaining predetermined values for the respective density data obtained in the obtained density data step. Micropattern shape inspection method characterized by having a defect determination step of determining acceptability of the fine pattern by comparing the constant has been quality criteria, the. 7. An image formed by detecting secondary electrons emitted from the measurement sample when the measurement sample is irradiated with an electron beam as image data to be inspected, the image data being inspected being curved in an uneven shape formed on the measurement sample. A micro pattern shape inspection apparatus for inspecting a minute micro pattern; an inspection area setting means for setting a curved inspection area corresponding to the shape of the micro pattern to the image data; Density peak point sequence means for obtaining a density peak point sequence based on the brightness distribution of a portion of the inspection area crossing the bending direction; and substantially perpendicular to the density peak point sequence obtained by the density peak point sequence means. Density data means for obtaining density data in a direction and in a range including the density peak points at a plurality of locations in the density peak point sequence; Defect determining means for calculating predetermined values for the respective density data thus obtained, and comparing these values with a preset quality determination criterion to determine the quality of the fine pattern. A micropattern shape inspection apparatus characterized in that: 8. The inspection area setting means has a function of setting the inspection area formed by each ellipse arranged outside and inside the edge of the micropattern with respect to the elliptical micropattern. The micropattern shape inspection apparatus according to claim 7, wherein: 9. The density peak point sequence means sets a straight line that intersects each ellipse of the inspection area at a predetermined angle, obtains each density peak position of each luminance distribution on the straight line, and obtains a density peak point sequence. The micropattern shape inspection apparatus according to claim 7, further comprising a function of acquiring a micropattern. 10. The acquisition density data means sets a range including the density peak point for acquiring the density data in accordance with an amount of unevenness of the minute pattern, an amount of distortion of the minute pattern, and a size variation thereof. 8. The micropattern shape inspection apparatus according to claim 7, wherein the apparatus has a function of performing the following. 11. The defect judging means adds luminance to each of the density data in a predetermined direction, and calculates maximum, minimum, average, and variance of the density addition value to obtain the maximum, minimum, and minimum values. 8. The micropattern shape inspection apparatus according to claim 7, further comprising a function of comparing the average, the variance value, and a preset quality judgment criterion to judge the quality of the micropattern. 12. A minute pattern shape inspection method for inspecting the minute pattern based on image data obtained by photographing a minute pattern formed in a concave and convex shape, wherein the minute pattern corresponds to a shape of the minute pattern. A first shape processing step of performing a filtering process on a density peak point sequence obtained based on the luminance distribution in the inspection area; A second shape processing step of performing a filtering process on the obtained density peak point sequence, a density peak point sequence filtered by the first and second shape processing steps and a density peak point sequence to be a reference. An evaluation step of comparing and evaluating the minute pattern. 13. The first and second shape processing steps include: an inspection area setting step of setting a curved inspection area corresponding to the shape of the fine pattern or a reference fine pattern for the image data. A density peak point sequence step of obtaining the density peak point sequence based on a luminance distribution of a portion of the inspection area crossing the bending direction set in the inspection area setting step; and the density peak point sequence obtained in the density peak point sequence step. 13. The micropattern shape inspection method according to claim 12, further comprising: a filtering step of performing a filtering process on the density peak point sequence. 14. The method according to claim 13, wherein in the filtering process, the density peak point sequence is subjected to Fourier series expansion and the shape is restored using Fourier coefficients up to a predetermined order. . 15. Image data obtained by detecting secondary electrons emitted from the measurement sample when the measurement sample is irradiated with an electron beam is curved in an uneven shape formed on the measurement sample as image data to be inspected. In a micropattern shape inspection apparatus for inspecting a minute micropattern, a first shape processing for filtering a density peak point sequence obtained based on a luminance distribution in an inspection area according to the shape of the micropattern Means, a second shape processing means for performing a filtering process on a density peak point sequence obtained based on a luminance distribution in an inspection area according to a shape of a reference minute pattern, and a first and a second processing means. Evaluation means for comparing the density peak point sequence filtered by the shape processing means with the density peak point sequence to be a reference to evaluate the minute pattern. A micropattern shape inspection apparatus. 16. The inspection area setting means for setting a curved inspection area corresponding to the shape of the fine pattern or a reference fine pattern for the image data, the first and second shape processing means. A density peak point sequence means for obtaining the density peak point sequence based on a luminance distribution of a portion of the inspection area that crosses the bending direction set by the inspection area setting means; and the density peak point sequence means obtained by the density peak point sequence means. 16. The fine pattern shape inspection apparatus according to claim 15, further comprising: a filter processing unit configured to perform a filtering process on the density peak point sequence. 17. The micropattern according to claim 16, wherein the filter processing means has a function of expanding the density peak point sequence into a Fourier series and restoring the shape using Fourier coefficients up to a predetermined order. Shape inspection device. 18. The function of adjusting the evaluation means so that at least the width of each pattern formed by the filtered density peak point sequence and the reference density peak point sequence is at least the same. 16. The micropattern shape inspection apparatus according to claim 15, comprising: 19. The evaluation means superimposes the filtered density peak point sequence and the reference density peak point sequence, and calculates each of the density peak point sequence and the reference density peak point sequence. The micropattern shape inspection apparatus according to claim 15, further comprising a function of setting a sum of shift amounts between patterns as a shape distortion evaluation value of the micropattern. 20. An image signal obtained by photographing a minute pattern, determining each edge on both sides corresponding to a contour of the minute pattern, and determining a line width of the minute pattern from a distance between the edges. Characteristic micro pattern shape inspection method. 21. A fine pattern shape inspection apparatus for inspecting the fine pattern based on an image signal obtained by detecting secondary electrons emitted from the measurement sample when the measurement sample is irradiated with an electron beam, A smoothing means for smoothing a waveform with a plurality of smoothing sizes for the image signal obtained by photographing the minute pattern; and smoothed waveform data having different smoothing sizes obtained by the smoothing means. Difference waveform means for calculating the difference waveform data of the difference waveform data obtained by the difference waveform means and the smoothed waveform data used by the difference waveform means for obtaining the difference waveform data. Width measuring means for determining each edge on both sides corresponding to the contour portion of the pattern, and determining the interval between these edges as the line width of the fine pattern. A micropattern shape inspection apparatus characterized by the following. 22. A regression curve is obtained from a series of line widths of the fine pattern obtained with the smoothed size that is equal to or more than the blur amount due to convolution by the diameter of the electron beam, and the smoothed size of the regression curve is zero. 22. The fine pattern shape inspection apparatus according to claim 21, further comprising a function of calculating a line width of the fine pattern when the diameter of the electron beam flux is zero from a position value.