Recently, the proportion of MI in the semiconductor indus-
try is growing rapidly due to the semiconductor yield issue.
It is important to develop, increase yield, and reduce time in
semiconductor process, and the key for this is MI. Thin-Flim
Metrology is a measurement equipment that is primarily used
in semiconductor casting processes during the MI process.
The Thi-Flim Metrology (thin film thickness measurement)
of the photolithographic to deposition process is one of the
most important element technologies because if the wafer
Vision-Alignmen is defective, numerous defective wafers can
be produced due to incorrect measurement. Accordingly, it
is necessary to accurately find and align the variable wafer
and die patterns using the shape pattern finding algorithm and
Ellipsometer position alignment technology acquired from the
wafer As the pattern becomes finer and high-degree stacking
is repeated, even a small distortion of the alignment leads to
defects.
It is essential to localize advanced stage and alignment
algorithms for Ellipsometer analysis. Stage and Align algo-
rithms linked to Vision/Elipsometer that can accurately specify
Wafer’s location are essential for the development of semicon-
ductor production processes and inspection of fine patterns
and limit stacks The Ellipsometer measures the change in
polarization state after reflection or transmission of light, and
the change in polarization state is measured It is a measuring
instrument that is determined by the characteristics of the sam-
ple (thickness, complex refractive index, or dielectric function)
and has a resolution of up to several angstroms (1 ˚
A = 1.0
x 10–10 m = 0.1 nm). Therefore, each company is carrying
out fine pattern identification and recognition of problems
during the process by applying the Ellipsometer production
process, and is promoting productivity maximization through
process improvement activities. For this, the interlocking Align
algorithm, which is HW Wafer Stage and SW, is essential, but
the technology is dependent on foreign products. Existing fine
pattern recognition methods were only able to align through
designated marks, but in this topic, we developed fine pattern
recognition-based align s/w through the form of repetitive
patterns. Previously, a separate space was needed for Fiducial
mark, but this study does not require a separate space for
Pattern Wafer x/y Auto Align System using Machine Vision
1TAE-YONG KIM, 1JONGPIL JEONG, 1CHAE-GYU LEE, 1SEONGJIN OH, 1LEE JIEUN, 2YONGJU NA
1Department of Smart Factory Convergence, Sungkyunkwan University, Suwon-si, REPUBLIC OF KOREA
2AI Research Lab, AIM, Hanam-si, REPUBLIC OF KOREA
Abstract: The paper proposes an Automatic Semiconductor Measurement System using Wafer Auto Align using
Pattern for semiconductor wafer measurement. The measurement of semiconductors is crucial for the
semiconductor industry, and the proposed model aims to improve the semiconductor production automation
process. The proposed system consists of three main components: the stage, the vision system, and the pattern
alignment algorithm. The stage includes theWafer holder, Ellipsometer, and controller, and plays a critical role
in aligning the X and Y axes of the Wafer to 100 mm/s after pattern analysis. The vision system captures high-
quality images of the Wafer and analyzes the patterns on the Wafer to detect any defects or deviations from the
standard. The pattern alignment algorithm uses the information obtained from the vision system to align the
Wafer accurately. The Auto align process is fully automated and does not require any user intervention. The
process operates in three major steps: selecting the Wafer Recipe, photographing the pattern of the designated
recipe, and executing the Auto align. The proposed system offers a comprehensive and automated solution for
Wafer alignment and measurement, providing high accuracy and efficiency, while also reducing the risk of
errors and improving the semiconductor production process.
Keywords: vision, ellipsometer, semiconductor, pattern alignment algorism, align
Received: May 18, 2022. Revised: June 18, 2023. Accepted: July 16, 2023. Published: August 3, 2023.
1. Introduction
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2023.19.6
Tae-Yong Kim, Jongpil Jeong,
Chae-Gyu Lee, Seongjin Oh, Lee Jieun, Yongju Na
E-ISSN: 2224-3488
53
Volume 19, 2023
Align If the pattern registration work to be a reference point
is carried out, it will not affect Align even if the production
processor or pattern is the same.
The paper proposes an Automatic Semiconductor Mea-
surement System using Wafer Auto Align using Pattern for
semiconductor wafer measurement. The system consists of a
stage, a vision system, and a pattern alignment algorithm.
The stage aligns the X and Y auto align to 100mm/s after
pattern analysis, while the vision system captures and analyzes
patterns. The pattern alignment algorithm processes the pattern
image and commands the stage controller’s auto alignment.
The process does not require user intervention, and the results
can be checked through UI/UX. The proposed model supports
various wafer pattern recipes and can be used for different
measurement processes without user intervention. The system
aims to improve semiconductor production processes and
increase productivity and profits for semiconductor companies.
The composition of this paper consists of five chapters as
follows. Section 1 describes the background and necessity
of the study as an introduction, and Section 2 introduces
Ellipsometer, Semiconductor Align, and Pattern alignment
algorithm necessary for this study. Section 3 describes the
structure, process, and overall layout of the proposed model
Section 4 describes the experimental environment, model
implementation, and quantitative evaluation of the proposed
model. The last section 5 summarizes the conclusion and
proposal models and describes future research plans.
Ellipsometry is technically complex as it is used in a
huge number of applications compared to other equipment,
so there are various types of lipometers suitable for various
applications, mainly in research institutes and semiconductor
industries. Among the various types of hi1ps00e161 are Rotat-
ing Polarizer Elipsometry (RPE), Rotating Ana-lyzer Elipsom-
etry (RAE), and Rotating Compenser Elip-someryRCE (S).
Eipsometry has a process that must be done before making a
measurement. Because it is a light measurement technology,
the light used in 806 must be well aligned with each opti-
cal component and the specimen to be measured, which is
called a18n.0ema. After alignment, optical components such
as po0larizC09 and analyzer have an optical axis, so it is
necessary to find an incident surface that changes slightly
every time the specimen is placed, and the position angle of
the optical components is called ali5ra660n.In most cases, the
calibration time is much longer than the actual measurement
time. In the calibration process, as the number of optical
components increases depending on the type of ellipsometer,
the position angle to be found increases accordingly. By
using the compen-sator, a phase delay compensation plate,
the experimental error that occurs when the reflected light
approaches linear polarization can be reduced, and since the
position is fixed while the polarizer ana-lyzer obtains data,
there is no residual polarization or polarization sensitivity
problem of the detector. In other words, it is elipsometry that
eliminates the shortcomings of RPE and PAE. However, the
calibra-tion process is much more complicated than RPE and
RAE because it is necessary to find the position angle of
the compressor as well as the polarizer and the analyzer. For
this reason, ordinary users recognize the Ellipsometer as an
equipment that is difficult to use.[1]
Fig. 1. Ellipsometry
Wafer chips manufactured through semiconductor process-
ing are used as key components of electronic devices in
various industries. There are also various types of wafer
defects or defects that occur when passing through these
various process processes [3]. Stacking patterns formed on
each layer vertically and continuously without missing the
correct position is called an overlay. Accurate alignment
techniques are required as one of the ways to increase overlay
values [4]. The problem of misalignment can be minimized
by increasing alignment technology that establishes the circuit
of the mask to be newly formed in the circuit formed on
the wafer and precisely adjusts the X and Y values. If the
position value to be devised when stacking circuits vertically
is an overlay, there is a critical dimension that horizontally
represents the uniformity of circuits. This is the distance
between the patterns and the minimum line width, and the
CD value should not vary depending on the location of the
wafer. In this way, the correction value should be calculated
using the overlay result and the calculated value should be
fed back to the exposure equipment to prevent misalignment
from occurring on subsequent wafers. In addition, methods
and devices for measuring errors in each unit process are
being actively studied. As semiconductor devices become
highly integrated in the photo process, accurate alignment
can become difficult and problematic depending on issues
such as alignment margin reduction, level stacking structure,
and wafer Daegu hardening. In addition, equipment such as
wafer stage, plate stage, lens, etc., and various defective issues
in design can also affect the misseline problem. Alignment
is one of the machine vision technologies that uses camera
sensors to recognize and calibrate the position of alignment
marks. Machine vision systems use special optical devices to
collect images with digital sensors that are protected inside the
camera, allowing the computer system to process and measure
various characteristics for decision making[5, 6]. Image and
2. Related Work
2.1 Ellipsometer
2.2 Semiconductor Align
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2023.19.6
Tae-Yong Kim, Jongpil Jeong,
Chae-Gyu Lee, Seongjin Oh, Lee Jieun, Yongju Na
E-ISSN: 2224-3488
54
Volume 19, 2023
image processing technologies capable of high resolution and
high accuracy are increasing, and recent deep learning imaging
using technologies such as image, lidar, ultrasound, and laser
is gaining research value among computer vision and image
processing. [7]
Visual object tracking is a fundamental task in the field
of computer vision. There are many applications such as
video surveillance, human computer interaction, traffic pattern
analysis, and robotics. Typical visual trackers can be classified
in two types of ways [8]. One method is target representation
and localization to cope with changes in the shape of the
target. Other methods are filtering and data connection, which
address the dynamics of tracked objects, scene pre-learning,
and evaluation of other hypotheses. Formulation of filtering
and data connection processes is achieved through a state space
approach for modeling discrete-time dynamic systems [3]. If
the dynamic and measurement functions are linear and the
noise sequence is Gaussian, Kalman filters provide the optimal
solution [4]. For various tracking scenarios, we applied the
aforementioned filtering and association methods to computer
vision. In this paper, we propose a novel tracking algorithm
that can simultaneously overcome difficulties associated with
rapid lighting changes, partial occlusion, similar color back-
grounds, and low illumination. For the proposed tracking
algorithm, we introduce a binary pattern-based SBP model
consisting of several sets of SBPs. In addition, I proposed
a kernel-based similarity measurement between the two SBP
models for target localization. In addition, binary pattern-
based SBP models provide better identification in situations of
similar color regions or low illumination, where color-based
models tend to fail to track targets.[9]
Fig. 2. Alignment using patterns
This paper proposes an Automatic Semiconductor Measure-
ment System using Wafer Auto Align using Pattern for semi-
conductor Wafer measurement. The measurement of semicon-
ductors is becoming a very important field in the semicon-
ductor production automation process. The share of MI in
the semiconductor industry was about 10%, but it is growing
rapidly due to the recent semiconductor yield issue. Due to
competition to refine semiconductor processes, development,
increase in yield, and reduce process time are factors that max-
imize productivity and profits of semiconductor companies,
and MI process is a key solution for this. Thin-Flim Metrology
is a measurement equipment used in major semiconductor
processes during the MI process. Thin-Flim Metrology in
the exposure/deposition process is one of the most important
element technologies because it can produce numerous defec-
tive Wafer due to incorrect measurement if the Wafer vision
alignment is poor. Therefore, it is necessary to accurately find
and align the highly variable Wafer and die patterns using
the shape pattern finding algorithm and Ellipsometer Position
Alignment technology acquired from Wafer. Stage and Align
algorithms linked to Vision/Elipsometer, which can accurately
specify Wafer’s location, are essential to presuppose the de-
velopment of semiconductor production processes with fine
patterns, marginal stacks, and core equipment for inspection.
The following Fig.3 is the overall system configuration of
the proposed model.
Fig. 3. System Configuration Diagram
The proposed Automatic Semiconductor Measurement Sys-
tem using Wafer Auto Align using Pattern is an advanced
technology that aims to improve the semiconductor production
automation process. The measurement of semiconductors is a
critical aspect of the semiconductor industry, and the proposed
model aims to maximize productivity and profits of semicon-
ductor companies by optimizing the MI process.
2.3 Pattern Alignment Algorism
3. Proposed Idea
3.1 Automatic Semiconductor Measurement
System Composition Chart
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2023.19.6
Tae-Yong Kim, Jongpil Jeong,
Chae-Gyu Lee, Seongjin Oh, Lee Jieun, Yongju Na
E-ISSN: 2224-3488
55
Volume 19, 2023
The system configuration of the proposed model consists
of three main components. The first component is the stage,
which includes the Wafer holder, Ellipsometer, and controller.
The stage plays a crucial role in aligning the X and Y axes of
the Wafer to 100 mm/s after pattern analysis. The Ellipsometer
is used to measure the thickness of the thin film on the Wafer,
while the controller controls the movements of the stage during
the measurement process.
The second component of the system is the Vision System,
which includes cameras, lenses, and lighting. The Vision
System captures high-quality images of the Wafer and analyzes
the patterns on the Wafer to detect any defects or deviations
from the standard. The lighting used in the Vision System is
carefully designed to ensure that the images captured are of
high quality and suitable for analysis.
The third component of the system is the pattern alignment
algorithm, which is an essential part of the Auto align process.
The pattern alignment algorithm uses the information obtained
from the Vision System to align the Wafer accurately. The
Auto align process is fully automated and does not require
any user intervention. The results of the Auto align process
can be viewed in real-time through the system’s user interface.
Fig. 4. Proposed Model Configuration Diagram
After analyzing the pattern image of the Vision System
through the pattern alignment algorithm, the stage controller’s
control algorithm is commanded to perform auto alignment. As
shown in Fig.4, the proposed model is performed without the
user’s special system operation in the process of auto aligning
the X and Y axes, and the user can check the corresponding
results through UI/UX.
The following Fig.5 is the flowchart of the proposed model
process and operates in three major steps.
The process of selecting the Wafer Recipe is an important
step in the proposed system, as it determines the type of
pattern and the size of the Wafer, and the measurement point
and pattern shooting location are changed accordingly. This
Fig. 5. Progress flow chart
ensures that the system can adapt to various Wafer patterns
and sizes, and allows users to choose the Wafer Recipe they
want based on their specific requirements. Once the Wafer
Recipe has been selected, the system uses an Auto align
algorithm to photograph the pattern of the designated recipe.
The location of each of the four patterns designated according
to the Wafer Recipe type can be found, photographed, stored,
analyzed, and digitized to allow users to check in real-time on
the user interface. After the main stage of the Auto alignment
algorithm, the system tracks and re-shoots the pattern taken
in accordance with the input of the new Wafer and the
movement of the Wafer Stage. The Auto alignment algorithm
then calculates the error of the changed X/Y axis alignment
and attempts to align the Auto alignment of the Wafer’s X/Y
axis. This entire process is automated and does not require
any user intervention, and the process can be checked in real-
time through the user interface. The corresponding result value
is then transferred to the Ellipsometer Controller, and the
Auto align is executed. The process is repeated at the end
of Wafer’s measurement. The time required for each Wafer
recipe may vary depending on the number of measurements
and the complexity of the pattern, but the system is designed
to handle this variability and provide accurate and efficient
results. Overall, the proposed system offers a comprehensive
and automated solution for Wafer alignment and measurement.
By allowing users to select the Wafer Recipe and automat-
ing the alignment and measurement process, the system can
provide high accuracy and efficiency, while also reducing the
risk of errors and improving the overall productivity of the
manufacturing process.
3.2 X/Y Auto Align Process
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2023.19.6
Tae-Yong Kim, Jongpil Jeong,
Chae-Gyu Lee, Seongjin Oh, Lee Jieun, Yongju Na
E-ISSN: 2224-3488
56
Volume 19, 2023
The experimental environment for the dataset includes a
C++ programming language and the tensorflow or pytorch ma-
chine learning frameworks. The hardware used includes an In-
tel i9-10900K CPU, a ROBO-8115VG2AR-Q470 SBC board,
8GB DDR4 RAM (2ea), Samsung 870 EVO 1TB SSD (2ea),
K-RACK 5020TL 2,5” 2BAY, PBPR-12P4 BACKPLANE,
RMC-4S 19” CHASSIS, DELTA GPS-1300CB POWER, and
a 1.5U CPU cooler. The development environment and ver-
sion are critical when using vision inspection and associated
libraries (MILLs).
In order to construct the layout of the proposed model, size
information and pattern information of each Wafer Recip are
required. The following Fig.6 summarizes the size information
and measurement data of Wafer Recip, which are essential data
to be included in the layout composition.
Fig. 6. Wafer recipe
The number of points measured for each Wafer size is
different, and this paper will use the most commonly used
41 Point Wafer Recip.
Fig. 7. Wafer recipe data
The following Fig.7 summarizes the details of Wafer Recip,
which is essential data to be included in the layout configura-
tion, and the Wafer Recip that you want to measure using the
data can be registered and used.
The following Fig.8 is a UI/UX screen that allows users
to analyze and measure the pattern after obtaining the Wafer
Pattern Image using the Image acquisition function, and check
the accuracy of the Auto Align.
Fig. 8. Auto align UI/UX
The Wafer Pattern Image output on the left side of the
screen is an image view for calculating the angle of the
changed X/Y axis with the image taken and obtained before
measurement, and the information of the corresponding X/Y
axis is listed below. The Wafer Pattern Image printed on the
right is an image obtained by re-taking the newly introduced
Wafer Pattern after measurement, and is configured to compare
it with the image taken before measurement, and information
on the changed X/Y axis below the image is listed. By
calculating the error of the changed X/Y axis through Auto
alignment algorithm, the align of the changed X/Y axis is
aligned and configured so that the user can check through the
corresponding UI/UX.
Fig. 9. Before proceeding with the pattern sorting algorithm
The following Fig.9 is a screen that sets the area of Wafer
Pattern after obtaining an image and proceeds with Pattern
Alignment Algorithm using data from the existing Wafer
Recip.
4. Experiment and Results
4.1 Data Description and Environment
Description
4.2 Configuring Layouts
4.3 Auto Align After Pattern Analysis and
Recognition (Model Implementation)
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2023.19.6
Tae-Yong Kim, Jongpil Jeong,
Chae-Gyu Lee, Seongjin Oh, Lee Jieun, Yongju Na
E-ISSN: 2224-3488
57
Volume 19, 2023
Fig. 10. After the progress of the pattern sorting algorithm
The following Fig.10 shows the result of auto alignment
through pattern alignment algorithm, and the result of final
alignment completion is indicated in pink area.
The first task is to develop a GUI that can continuously
validate the integration of the Align algorithm with the vision
system. The second task is to develop iterative algorithms for
machine vision results and fine capture, implement alignment
algorithms to achieve 99percent accuracy, and are validated.
The following Table 1 is a table of Tact Time.
TABLE I
TACT TIME
1st 2nd 3rd 4th 5th
TACT Time 14 15 15 14 16
The proposed technology presents a promising solution for
achieving accurate alignment in a production setting while
offering several benefits compared to existing methods. The
technology’s primary advantage is the ability to provide similar
performance as traditional alignment methods but with reduced
tact time and the elimination of the need for fiducial marks.
Traditional alignment methods require the use of fiducial
marks, which are reference points used to align parts or com-
ponents. The process of placing and identifying these marks
can be time-consuming, leading to increased tact time, and
the marks may not be reliable due to shifting or misalignment
during production. In contrast, the proposed technology elim-
inates the need for fiducial marks, streamlining the alignment
process and significantly reducing the tact time required.
Another benefit of the proposed technology is the user-
friendly GUI, which allows for continuous validation of the
integration of the Align algorithm with the vision system. The
GUI provides a simple and efficient way to monitor and adjust
the technology as needed, ensuring that it is always functioning
correctly and providing accurate results.
In summary, the proposed technology offers a reliable
and efficient solution for achieving accurate alignment in a
production setting, with the added benefits of reduced tact
time and the elimination of the need for fiducial marks.
The technology achieves this through the use of iterative
algorithms, integration of alignment algorithms, and a user-
friendly GUI, making it a promising alternative to traditional
alignment methods..
In this paper, the authors proposed a pattern matching
alignment method using computer vision technology. The goal
of the method was to align objects with high accuracy and
efficiency in manufacturing processes, such as semiconductor
wafer processing. To evaluate the performance of the proposed
method, the authors conducted experiments and measured the
error rate and tact time. The error rate is a measure of how
accurately the method can align the objects, and the tact
time is a measure of how quickly the method can perform
the alignment. The results of the experiments showed that
the proposed method achieved an error rate of within 5%,
which indicates that it can align objects with high accuracy.
Moreover, the tact time was faster than other methods, which
suggests that the proposed method can improve the efficiency
of manufacturing processes. Overall, the results of the ex-
periments demonstrated that the pattern matching alignment
method using vision proposed in the paper is a promising
approach for achieving accurate and efficient object alignment
in manufacturing processes. The authors believe that their
method can have significant practical applications in various
industries and can contribute to the development of advanced
manufacturing technologies.
In this paper, a model of pattern alignment algorithm using
Pattern Wafer’s pattern was proposed. Based on the technology
of the model, we will also develop an Auto align measurement
model of plain Wafer using Vision to reproduce the complete
Automatic Semiconductor Measurement System Composition
chart.
Following are results of a study on the ”Leaders in IN-
dustryuniversity Cooperation 3.0” Project, supported by the
Ministry of Education and National Research Foundation of
Korea, Corresponding author: Prof. Jongpil Jeong
[1] Jae-Sun Kyung, Hye-Keun Oh, Il-Shin Ahn, and Ok-Kyung Kim.
”Calibration and Incidence Plane Fixed Ellipsometer in Ellipsometry.
Journal of the Korean Society of Semiconductor Equipment 2.4 (2003).
pp: 23-26.
[2] Ju-Yong Park. ”A Study on a Model Combining Super-resolution Based
Attention Mechanism for Alignment Mark Detection in Semiconductor
Optical Equipment. D. thesis, Sungkyunkwan University, Graduate
School of General Studies, 2022.
[3] Kim, Sungho, (2016) ”Real-time 3D human whole body fusion using a
kinect Motion Capture Using Kinect Sensor”, Korean Society for Digital
Policy, Digital Convergence Research, Vol. 14, No. 1, pp. 189-194
4.4 Result
5. Conclusion
Acknowledgment
References
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2023.19.6
Tae-Yong Kim, Jongpil Jeong,
Chae-Gyu Lee, Seongjin Oh, Lee Jieun, Yongju Na
E-ISSN: 2224-3488
58
Volume 19, 2023
[4] In-Joong Kim, (2007) ”Adaptive binarization for camera-based doc-
ument recognition”, Proceedings of the Korea Industrial Information
Society, Vol. 12, No. 3, pp. 132-140
[5] Lee, Seok-Ki, Kim, Seok-Tae, and Cho, Sung-Jin, (2003) ”Noise re-
moval and edge extraction of images using Potts Automata”, Journal
of the Korean Communications Society, Communication Theory and
Systems, Vol. Vol. 28, No. 3C, pp. 327-335
[6] Lee, Jungseop, (2009), ”Automatic classification algorithm of semicon-
ductor package mold surface defects using pattern recognition”, Master’s
thesis, Hoseo University Graduate School
[7] Minchul Kim. ”A Study on Alignment Model for Improving Semicon-
ductor Package Inspection Efficiency. Master’s thesis, Gongju Univer-
sity, Graduate School of General Studies, 2019. Chungcheongnam-do,
South Korea
[8] Kim, S., Kim, D. H., Choi, S. H., Kim, W. Y., Kwon, S., &Cho, Y.
T. (2020). Effect of surface pattern morphology on inducing superhy-
drophobicity. Applied Surface Science, 513, 145847.
[9] Kim, Woo-Young. ”A Study on Precision Alignment System using Mi-
cro Pattern Characteristics. D. Thesis, Changwon National University,
2021. Gyeongsangnam-do, South Korea
[10] Liu Image Retrieval on Real-Life Image With Pre-Trained Vision-and-
Language Models ICCV 2021 paper
Contribution of Individual Authors to the
Creation of a Scientific Article (Ghostwriting
Policy)
The authors equally contributed in the present
research, at all stages from the formulation of the
problem to the final findings and solution.
Sources of Funding for Research Presented in a
Scientific Article or Scientific Article Itself
Conflict of Interest
The authors have no conflicts of interest to declare
that are relevant to the content of this article.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en
_US
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2023.19.6
Tae-Yong Kim, Jongpil Jeong,
Chae-Gyu Lee, Seongjin Oh, Lee Jieun, Yongju Na
E-ISSN: 2224-3488
59
Volume 19, 2023
Following are results of a study on the ”Leaders in IN-
dustryuniversity Cooperation 3.0” Project, supported by the
Ministry of Education and National Research Foundation of
Korea, Corresponding author: Prof. Jongpil Jeong