KR102171206B1 - Method for quantitative measurement of defects within block-copolymer thin-flim having lamellar pattern - Google Patents

Abstract

The present invention relates to a method for analyzing a region affected by morphological defects in a block-copolymer thin film having a self-assembled lamella pattern, including a series of processing steps using SEM images.

Description

{Method for quantitative measurement of defects within block-copolymer thin-flim having lamellar pattern}

The present invention relates to a method for analyzing a region affected by morphological defects in a block-copolymer thin film having a self-assembled lamella pattern, including a series of processing steps using SEM images.

Memory devices are typically provided as internal semiconductor integrated circuits within a computer or other electronic device. Memory devices usually have a number of different types of memory, including random access memory (RAM), read only memory (ROM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), and flash memory. Flash memory devices have evolved into a popular source of nonvolatile memory for a wide range of electronic applications. Flash memory devices typically have to take into account high memory density (density), high reliability and low power consumption.

Since the development of the bipolor transistor at Bell Labs in 1947, the degree of integration of semiconductor chips has evolved according to Moore's law. These developments have been faithfully maintained thanks to the continuous development of the optical etching process, but according to the recent ITRS (International Technology Roadmap for Semiconductor), it is known that process technology of 30 nm or less has many problems to implement due to the limitation of optical etching technology . Accordingly, various nanopattern fabrication technologies are being developed based on new principles, and among these research fields, research on molecular assembly nanostructures is receiving worldwide attention.

As described above, attempts to increase the degree of integration of semiconductor devices in order to increase the memory density in semiconductor devices such as flash memories are continuing. Planarly, the area occupied by each unit cell has been reduced, and in response to such a decrease in unit cell area, a design rule of a smaller nanoscale CD (Critical Dimension) of several to several tens of ㎚ is applied. Accordingly, a new technique for forming a fine pattern such as a fine contact hole pattern having a nano-scale opening size or a fine line pattern having a nano-scale width is required.

When relying only on top-down photolithography technology to form fine patterns for semiconductor device manufacturing, there is a limitation in improving resolution due to the wavelength of the light source and the resolution limit of the optical system. . As one of efforts to overcome the limitation of resolution in the photolithography technology and develop a next-generation microfabrication technology, methods of forming a microstructure in a bottom-up method using a self-assembly phenomenon of molecules have been attempted.

Molecular assembly The most representative polymer material that forms nanostructures is a block copolymer (BCP), which has a molecular structure in which chemically different unit blocks are connected through covalent bonds. Therefore, the block copolymer can form various nanostructures such as spheres, cylinders, and lamellas having a regularity of several to tens of nanometers, and is thermodynamically stable and has nanostructures. It has the advantage of being able to design size and physical properties at the synthesis stage. In addition, it is possible to quickly implement nanostructures in a large area through a parallel process, and it is easy to remove the template after forming an inorganic and organic nanostructure thin film using a block copolymer template, nanowires, quantum dots, magnetic storage media, nonvolatile memory In the fields of IT, BT, ET, etc., intensive research is underway with nanopattern manufacturing technology for manufacturing various next-generation devices.

The lamellar pattern that appears from the double self-assembled block copolymer thin film is a promising sub-10 nm nanoscale pattern for next-generation devices such as memory, storage and arithmetic logic units (ALU). It is the material. Quantitative measurement of the quality of the lamella pattern is an important factor determining whether or not the developed pattern can be applied to an actual device. However, the conventional measurement method is dependent on an optical microscope that detects local defects by image processing, and it is difficult to provide detailed information on the quality of the pattern due to the technical limitations of the measurement method. Accordingly, there is a need for a quantitative measurement method capable of analyzing microscopic order, quality and uniformity of patterns and delivering specific information.

The present inventors have worked hard to find a treatment process capable of detecting defects in a self-assembled block copolymer thin film without complicated calculations through several steps of treatment using the original SEM image of the thin film surface to be inspected. The image entropy value is calculated through fast Fourier transform (FFT), conversion to a binary image, and Bragg filtering, and the density of defects as well as the area ratio of the pattern affected by the inverse FFT reconstruction image entropy are calculated. It was confirmed that it was possible, and the present invention was completed.

A first aspect of the present invention is a method for analyzing a region affected by morphological defects in a self-assembled block copolymer thin film, comprising: a first step of preparing an original SEM image of the surface of the self-assembled block copolymer thin film; A second step of converting the original image into a gray-scale image; A third step of deriving a fast Fourier transform (FFT) image of the gray-scale image by using a discrete Fourier transform algorithm; A fourth step of selecting a Bragg dot pattern from a 2D FFT image for a Bragg filter; A fifth step of applying a Bragg filter to the filtered image; A sixth step of calculating image entropy of the filtered image; A seventh step of calculating a 2D convoluted filtered image using image entropy; An eighth step of calculating the image entropy of the combined image; And it provides an analysis method comprising a ninth step of deriving the 2D combined image obtained from the steps 5 to 8.

A second aspect of the present invention is step A of preparing a self-assembled block copolymer thin film; Step B of performing the analysis method according to the first aspect to analyze a region affected by morphological defects in the prepared self-assembled block copolymer thin film to calculate an area fraction of the defect; And it provides a quality control method of a self-assembled block copolymer thin film comprising the step C of selecting a product in which the calculated area fraction of the defect is 0.5% or less of the total area of the thin film.

A third aspect of the present invention is the step I of preparing a self-assembled block copolymer thin film having a regular pattern in which the area fraction of the defects selected by the method of the second aspect is 0.5% or less of the total area of the thin film; And an II step of transferring the self-assembled block copolymer thin film having a regular pattern onto a desired material as a template.

Hereinafter, the present invention will be described in more detail.

The present invention is a method for analyzing regions affected by morphological defects in a self-assembled block copolymer thin film,

A first step of preparing an original SEM image of the surface of the self-assembled block copolymer thin film;

A second step of converting the original image into a gray-scale image;

A third step of deriving a fast Fourier transform (FFT) image of the gray-scale image by using a discrete Fourier transform algorithm;

A fourth step of selecting a Bragg dot pattern from a 2D FFT image for a Bragg filter;

A fifth step of applying a Bragg filter to the filtered image;

A sixth step of calculating image entropy of the filtered image;

A seventh step of calculating a 2D convoluted filtered image using image entropy;

An eighth step of calculating the image entropy of the combined image; And

It provides an analysis method comprising a ninth step of deriving the 2D combined image obtained from steps 5 to 8 above.

For example, the second step may be performed by applying a luminosity algorithm known in the art. In this case, the strength of the signal may be defined as 0.21R+0.72G+0.07B. However, the present invention is not limited thereto, and non-limiting methods known in the art may be performed alone or in combination to achieve the above object.

Specifically, the third step converts a mathematically uniform and complete 1D pattern having a single periodicity into a set of discrete values of Fourier coefficients, and converts the experimentally obtained 1D pattern of the second step. , By applying a fast Fourier transform (FFT), it can be achieved by expressing a diffraction pattern (FIG. 1B) such as a dot pattern in a reciprocal space.

)의 중첩에 의해 1D 원통형 패턴에 상응하는, 선택적으로 증강된(intensified) 신호(

)를 제공하는 맵을 구축함으로써 달성할 수 있다.Specifically, the fourth step is to scan the Fourier transformed amplitudes of the original SEM image by focusing on the dot pattern in the inverse lattice space, and the extracted amplitude (

), corresponding to the 1D cylindrical pattern by the superposition of, optionally intensified signal (

This can be achieved by building a map that provides ).

) 및 증강된 신호(

)의 관계는 하기 방정식 1로 표현될 수 있다:At this time, the extracted amplitude (

) And augmented signal (

) Can be expressed by Equation 1:

[Equation 1]

The k denotes a reciprocal wave vector, and P _k ( r ) denotes the phase at position r .

For example, in the fourth step, a circular-shaped scanning Gaussian function is set to sufficiently cover the Bragg spot, and the dot pattern in the FFT image is numerically scanned with a designated Gaussian filtering function, and a filter An inverse fast Fourier transformed image can be obtained by constructing an amplitude map in the inverse lattice space with the resulting amplitude.

Specifically, the fifth step is to convert the inverse fast Fourier transformed gray-scale image to a black and white image in order to enhance the filtered amplitude of the aligned 1D pattern, mean-square filtering, Alternatively, it can be performed by applying bilateral range filtering.

For example, in the process of the mean-square filter, the square window is an inverse fast Fourier transform with a shift-invariant Gaussian type filter having a reference signal value set as the average of signals obtained from inverse fast Fourier transformed images. The resulting image can be scanned and filtered in both directions (IEEE Int'l Conf. Comp. Vis., Bombay, India, 1997). Test three or four sets of window sizes and both patterns, select the combination to obtain the best filtered image, and compare to the bare Bragg filtered image, the mean-square filtered image is morphological in a 1D cylindrical pattern. It can exhibit highly reinforced and discernable forms of defects.

Specifically, S _i for calculating the image entropy of the i-th pixel of the image in the sixth step may be defined by Equation 2:

[Equation 2]

In Equation 2 above,

P _ik is the probability of an intensity difference between the i-th pixel and its k-th neighboring pixel.

Specifically, in the 7th step, the 2D-combined filtered image is a mean-square filtered binary image for generating an additionally processed binary image that selectively filters and excludes a region having a morphological defect. Acting as a convolution kernel, local entropy can be significantly increased when placed in defective regions adjacent to morphological defects, deflective orientations, or both.

Specifically, in step 8, another image entropy from the processed binary image is calculated by calculating pixel-by-pixel entropy with related entropy filtering neighbors by considering a specified number of neighboring pixels. This can be achieved by calculating the filtered image.

For example, morphological defects in a self-assembled block copolymer thin film are non-perfectly aligned dislocations and +1/2 or -1/2 disclinations present in specific regions of the image. It provides streaks, thus exhibiting non-zero texture randomness and can increase image entropy.

Specifically, the ninth step is to combine the filtered binary image of the 1D pattern and the image entropy filter, but the image entropy filter, S ( m , n ) (where m and n are pixel indexes) are processed images, and equation 3 below Accordingly, a filtered binary image to produce P ( m , n ), which is achieved by numerically counting the pixels in the defective region, can serve as a binding core for B ( m , n ):

[Equation 3]

In Equation 3 above,

J and K are the convolution kernel dimension.

More specifically, given the processed image P ( m , n ), a region affected by a morphological defect may be detected at a designated threshold value, and an area fraction of the defect may be calculated. .

For example, 5% of the maximum value of P ( m , n ) was set as the titer, and a region having a value less than that was selected to calculate its fraction.

In order to provide a device with high performance, it is necessary to establish a fine pattern process as the demands for miniaturization and density continue to continue. In particular, fabricating high-density nanopatterns over a large area through a low-cost process is an important factor in the development of various next-generation nanodevices. In the conventional semiconductor process, optical lithography techniques such as I-line and ArF were mainly used, but in order to manufacture a pattern at a level of 40 nm or less, a high-energy light source is required, and there are numerous technical difficulties in developing optical equipment according to it. , There is also a phenomenon in which the photoresist forming the pattern collapses. On the other hand, processes using high-energy light sources such as electron beam or extreme ultraviolet lithography can solve patterns of 20 nm or less, but are slow due to the formation of tandem patterns, and these disadvantages appear larger as the scale is reduced in a large area. . In addition, in the case of nanoimprint, which is another method, since the pattern is transferred through direct contact between the mask and the photoresist, contamination or damage of the mask occurs during this process, which is pointed out as a fatal problem.

Accordingly, a method of fabricating a nano pattern using a self-assembled nano structure formed by itself is attracting attention. Block copolymers, which are representative molecular assembly materials, have a size of about 5 to 50 nm because the ends of polymer chains with different chemical properties are connected by covalent bonds, and thus, spheres, cylinders, lamellars, etc. Spontaneously form periodic nanostructures of various types. This process is a parallel process that can be mass-produced. Since the nanostructure is thermodynamically stable, it is advantageous to fabricate a pattern having a high aspect ratio, and the size of the nanostructure can be variously controlled. Such a block copolymer nanostructure can be used as an etching and deposition template in a pattern transfer process. Therefore, block copolymer lithography is considered to have sufficient potential as nanolithography that can be applied to the fabrication of nanodots and nanowires, and to fabricate various next-generation nanodevices such as photonic crystals, high-performance magnetic storage media, large-capacity memory devices, semiconductor devices, and solar cells. .

Furthermore, it is important in the fabrication of functional nanostructures to form self-assembled block copolymer nanostructures on various substrates over a large area. However, in the case of a nanostructure formed by a naturally formed self-assembled block copolymer pattern, the structure is irregular and may contain many defective structures, so it is necessary to check the quality before practical use.

In this context, the present invention is a step of preparing a self-assembled block copolymer thin film;

B step of calculating the area fraction of the defect by analyzing the area affected by the morphological defect in the prepared self-assembled block copolymer thin film by performing the analysis method of the present invention described above; And

It provides a method for quality control of a self-assembled block copolymer thin film comprising the step C of selecting a product in which the calculated area fraction of the defect is 0.5% or less of the total area of the thin film.

Further, the present invention provides a step I of preparing a self-assembled block copolymer thin film having a regular pattern in which the area fraction of the defects selected by the above method is 0.5% or less of the total area of the thin film; And

It provides a method of manufacturing a semiconductor device comprising the step II of transferring the self-assembled block copolymer thin film having a regular pattern onto a desired material as a template.

For example, the self-assembled block copolymer thin film having a regular pattern may be manufactured by a graphoepitaxy technique, but is not limited thereto.

The analysis method of the present invention does not simply determine the type and/or number of defects, but calculates the area of the area affected by the actual defects, and does not include a complicated calculation process such as distinguishing the types of defects. Since it is possible to quickly and accurately determine the quality of a self-assembled block copolymer thin film without consuming resources, it can be used for quality control in the device preparation stage in a semiconductor production process based on this.

1 is a diagram showing a process of processing an original raw SEM image for quantitative detection of morphological defects in a self-assembled pattern of a BCP thin film. (a) is the original SEM image, (b) is the 2D FFT image of the original image, (c) is the Bragg-filtered image, including the inset of the lower left corner initiating the Bragg-filtered spot, ( d) is a mean-square filtered or two-sided black and white filtered binary image, (e) is an image entropy map of a mean-square filtered binary image, and (f) is a combination between the entropy map and the mean-square filtered binary image. The subsequent filtered binary image, (g) is the image entropy filter map of (f), and (h) is a map showing the final morphological defects (pink regions) created by the combination of (d) and (g). Represents.

Hereinafter, the present invention will be described in more detail through examples. These examples are for explaining the present invention more specifically, and the scope of the present invention is not limited to these examples.

Example One: Lamella Topological defects in patterns Of areal density Algorithm for analysis

In order to analyze the areal density of morphological defects in a block copolymer (BCP) thin film having three or four different types of morphological defects, a series of continuous processing and analysis using the raw SEM image of the BCP thin film as follows. The process was carried out:

Step 1. Prepare the raw raw SEM image,

Step 2. Convert the original image to a gray-scale image,

Step 3. Derive a 2D FFT image of the gray-scale image using a discrete Fourier transform (discrete cosine transform (DCT)) algorithm,

Step 4. Select Bragg dot pattern from 2D fast Fourier transform (FFT) image for Bragg filter,

Step 5. Applying the Bragg filter to the filtered image,

Step 6. Calculate the image entropy of the filtered image,

Step 7. Compute the 2D convoluted filtered image using image entropy to obtain the most-disordered local areas,

Step 8. Calculate the image entropy of the combined image, and

Step 9. Calculate the 2D combined images obtained from steps 5 to 8 above to obtain a final image representing the area affected by the defect in the lamella pattern.

-Fast Fourier Transform

After the application of shear leading to SVA (solvent vapor annealing), the nano-scale periodic structures of BCP are uniform line-and-space width. ), as well as uniform periodicity, showed an almost perfect 1D cylindrical pattern. Mathematically, a perfect 1D pattern with uniform and single periodicity was converted into a set of discrete values of Fourier coefficients. Experimentally, the 1D pattern (FIG. 1A) was expressed as a diffraction pattern (FIG. 1B) similar to a dot pattern in a reciprocal space. Coefficiently, a fast Fourier transform (FFT) can be applied to obtain the dot pattern of the SEM image experimentally observed for the 1D cylindrical pattern BCP thin film.

- Bragg Filtering

은, 다음과 같이, 가공하지 않은 SEM 이미지 신호(

)의 푸리에 합을 이용하여 얻을 수 있다:In order to obtain quantitative information on the 1D cylindrical pattern, Bragg filtering, which scans the Fourier transformed amplitudes of the original SEM image by focusing on the dot pattern in the inverse lattice space, can be applied. As in the context of the theoretical approach proposed by Hytch (Microsc. Microanal. Microstruct., 1997, 8: 41; J. Mater. Res., 2009, 24: 342), by superposition of amplitudes extracted from Bragg filtering. , It is possible to construct a map that provides selectively intensified signals corresponding to the 1D cylindrical pattern. The amplitude,

Is, as follows, the raw SEM image signal (

) Can be obtained using the Fourier sum of:

The k denotes a reciprocal wave vector, and P _k (r) denotes the phase at position r. For Bragg filtering, a dot pattern (inset of Fig. 1C) in the FFT image was counted and scanned with a designated Gaussian filtering function. A circular-shaped scanning Gaussian function was set to sufficiently cover the Bragg spot. With the filtered amplitude, an inverse fast Fourier transformed image could be obtained by constructing an amplitude map in the inverse lattice space (FIG. 1C).

-Mean-square Filtering (mean-square filtering, both ranges Filtering ; bilateral range filtering)

Bragg filtering provides a numerically reconstructed 1D cylindrical pattern that is different from background noise. Mean-square filtering, which converts the reconstructed (inverse fast Fourier transformed) gray-scale image to a black and white image to enhance the filtered amplitude of the aligned 1D pattern with minimized background noise. Filtering, or bilateral range filtering, was applied (Fig. 1d). In the process of the mean-square filter, the square window is an inverse fast Fourier transformed image with a shift-invariant Gaussian type filter with a reference signal value set as the average of the signal obtained from the inverse fast Fourier transformed image. Was scanned and filtered in both directions (IEEE Int'l Conf. Comp. Vis., Bombay, India, 1997). Three or four sets of window sizes and both patterns were tested, and combinations to obtain the best filtered image were selected. Compared to the bare Bragg filtered image, the mean-square filtered image showed a highly enhanced and discernable morphology of morphological defects in the 1D cylindrical pattern.

-Calculation of image entropy

Basically, a perfect 1D periodic pattern without any noise and any morphological defects represented a 1D periodic pattern of image entropy. To calculate the image entropy of the i-th pixel of the image, S _i can be defined as follows:

The P _ik represents a probability of occurrence of an intensity difference between an i-th pixel and its k-th neighboring pixel. Image entropy can act as a kernel for processing a binary image for quantitative detection of texture randomness around each pixel of a binary image. For example, higher pixel entropy corresponds to higher texture randomness (Fig. 1E). Therefore, the image entropy map in the form of a 2D matrix is a convolution kernel for a mean-square filtered binary image for the generation of an additionally processed binary image, which selectively filters and excludes regions with morphological defects. It can act as (Fig. 1f). This is mainly due to the fact that the local entropy increases significantly when the POI pixel is placed in a defective area, such as adjacent to morphological defects and deflective orientations.

-Image entropy Filtering

From the processed binary image, another image entropy filtered image could be calculated. To this end, a specified number of neighboring pixels were considered. As an associated entropy filtering neighbor, a filtered image entropy map was obtained by calculating pixel-by-pixel entropy (FIG. 1G).

-Detection of morphological defects

Morphological defects (ie, dislocations and +1/2 or -1/2 disclination) present in certain areas of the image can be found in non-perfectly aligned stripes as well as non-zeros. -zero) indicates texture randomness. For example, in the case of dislocation, there is a set of oblique line-and-space patterns around the dislocation. Also, the terminal point at the end of the dislocation significantly increases the image entropy. In the case of the +1/2 rotation, the end point and the curved morphology also increase texture randomness and improve image entropy. Also, in the case of the -1/2 turn, the contact point strongly modulates texture randomness and image entropy. Considering this characteristic of morphological defects in a 1D-arranged cylindrical pattern, a filtered binary image of a 1D pattern and an image entropy filter can be combined in order to quantitatively detect and calculate the area affected by the defect. In particular, the image entropy filter, S ( m , n ) (where m and n are pixel indices) is a processed image, a filtered binary image to produce P ( m , n ), combined for B ( m , n ) It can act as a key, and it numerically counts the pixels within the defective area as follows:

The J and K are the convolution kernel dimension. Given the processed image P ( m , n ), the area affected by the morphological defect was detected at a designated threshold value, and the area fraction of the defect was calculated (FIG. 1H ).

<Comparison with existing defect detection algorithms>

1. Conventional Algorithm

The conventional algorithm for detecting defects includes: a first step of securing an original fingerprint image, a second step of binarizing the image (black and white imaging); A third step of applying a block filter; A fourth step of tracking a ridge flow and calculating a crossing number; And a fifth step of detecting the details (minutiae).

-Depends on the existing fingerprint detect algorithm (minutiae detection).

-Count only the number of defects.

-Local bending, distortion, and width fluctuation cannot be detected.

2. The algorithm of the present invention

-Image entropy and convolution dependent

-Detect areas affected by defects as well as local warpage, bending and fluctuations.

3. Advantages of the algorithm of the present invention

-Provides more specific quantitative information on the quality of the pattern

-Defective areas within the pattern can be detected

-Fast & robust

Claims (15)

As an analysis method of a region affected by morphological defects in a self-assembled block copolymer thin film,
A first step of preparing an original SEM image of the surface of the self-assembled block copolymer thin film;
A second step of converting the original SEM image to a gray-scale image;
A third step of deriving a fast Fourier transform (FFT) image of the gray-scale image by using a discrete Fourier transform algorithm;
For a Bragg filter, a Bragg dot pattern is selected from a 2D FFT image, and the Fourier transformed amplitude of the original SEM image is scanned with a specified Gaussian filtering function by focusing on the Bragg dot pattern, and the Gaussian filtering function A fourth step of constructing an amplitude map with the amplitude scanned by and filtered;
A fifth step of applying a Bragg filter to an image including an amplitude filtered by the specified Gaussian filtering function;
A sixth step of calculating image entropy of the filtered image;
A seventh step of calculating a 2D convoluted filtered image using image entropy;
An eighth step of calculating the image entropy of the combined image; And
And a ninth step of deriving the 2D combined image obtained from the steps 5 to 8.
The method of claim 1,
The third step is to convert the mathematically uniform and complete 1D pattern with single periodicity into a set of discrete values of Fourier coefficients, and the experimentally obtained 1D pattern of the second step is converted to fast Fourier. The analysis method is achieved by applying a fast Fourier transform (FFT) and expressing it as a diffraction pattern (Fig. 1B) such as a dot pattern in a reciprocal space.
The method of claim 1,
The fourth step is to scan the Fourier transformed amplitudes of the original SEM image by focusing on the dot pattern in the inverse lattice space, and the scanned Fourier transformed amplitude (

), corresponding to the 1D cylindrical pattern by the superposition of, optionally intensified signal (

Analysis method that is achieved by building a map that provides ).
The method of claim 3,
The Fourier transformed amplitude (

) And the augmented signal (

) Is the analysis method that has the relationship of Equation 1:
[Equation 1]

The k denotes a reciprocal wave vector, and P _k ( r ) denotes the phase at position r .
The method of claim 3,
By setting a circular-shaped scanning Gaussian function to sufficiently cover the Bragg spot, the dot pattern in the FFT image is numerically scanned with a designed Gaussian filtering function, and the filtered amplitude in the inverse lattice space An analysis method to acquire an inverse fast Fourier transformed image by constructing an amplitude map.
The method of claim 1,
The fifth step is to convert the inverse fast Fourier transformed gray-scale image to a black-and-white image to enhance the filtered amplitude of the aligned 1D pattern, mean-square filtering, or both ranges. Filtering; Bilateral range filtering) is applied to the analysis method.
The method of claim 1,
In the sixth step, S _i for calculating the image entropy of the i-th pixel of the image is defined by Equation 2 below:
[Equation 2]

In Equation 2 above,
P _ik is the probability of an intensity difference between the i-th pixel and its k-th neighboring pixel.
The method of claim 1,
In step 7, the 2D-combined filtered image selectively filters and excludes regions with morphological defects, and the combining core for the mean-square filtered binary image for the generation of an additionally processed binary image ( convolution kernel), where local entropy is significantly increased when placed in a defective region adjacent to morphological defects, deflective orientations, or both.
The method of claim 1,
In the eighth step, another image entropy-filtered image from the processed binary image by calculating pixel-by-pixel entropy with the associated entropy filtering neighbors, taking into account a specified number of neighboring pixels. Analysis method that is achieved by calculating the.
The method of claim 1,
Morphological defects in the self-assembled block copolymer thin film are dislocations and non-perfectly aligned stripes that are +1/2 or -1/2 disclination in specific areas of the image. Provides, and thus non-zero (non-zero) texture randomness and to increase image entropy.
The method of claim 1,
In the ninth step, the filtered binary image of the 1D pattern and the image entropy filter are combined, but the image entropy filter, S ( m , n ) (where m and n are pixel indexes) are processed images, and are defective according to Equation 3 below. An analytical method that serves as a binding core for B ( m , n ), a filtered binary image to produce P ( m , n ), which is achieved by numerically counting the pixels in the region:
[Equation 3]

In Equation 3 above,
J and K are the convolution kernel dimension.
The method of claim 11,
Given the processed image P ( m , n ), an area affected by a morphological defect is detected at a designated threshold value, and an area fraction of the defect is calculated.
Step A of preparing a self-assembled block copolymer thin film;
A step B of calculating an area fraction of a defect by analyzing a region affected by a morphological defect in the prepared self-assembled block copolymer thin film by performing the analysis method according to any one of claims 1 to 12; And
A method for quality control of a self-assembled block copolymer thin film comprising step C of selecting a product having the calculated area fraction of defects of 0.5% or less of the total area of the thin film.
A step I of preparing a self-assembled block copolymer thin film having a regular pattern in which the area fraction of the defects selected by the method of claim 13 is 0.5% or less of the total area of the thin film; And
A method of manufacturing a semiconductor device comprising the step II of transferring the self-assembled block copolymer thin film having a regular pattern onto a desired material as a template.
The method of claim 14,
The method of manufacturing a semiconductor device in which the self-assembled block copolymer thin film having a regular pattern is manufactured by a graphoepitaxy technique.

Images