top of page

The key to outpacing China in semiconductors and displays is
'Artificial Intelligence' Plasma process capabilities greatly improved with AI

U.S. 'Lam Research' at the forefront of active adoption In Korea, major companies like Samsung and SK actively utilizing it… Ongoing research centered around government-funded institutes"

한국이 반도체·디스플레이 산업 선두 자리를 유지하기 위해선 첨단 공정 기술 연구가 끊임없이 이뤄져야 한다. 특히 전문가들은 ‘인공지능(AI)’ 기반 공정 기술 확보가 곧 미래 반도체·디스플레이 산업 주도권을 잡는 열쇠가 될 것이라고 강조한다./ 그래픽=박설민 기자

The South Korean Information and Communications Technology (ICT) industry is like a house of cards built on 'semiconductors' and 'displays'. According to the July report of this year on ICT import and export trends released by the Ministry of Science and ICT, Korea's export amounts for semiconductors and displays are $7.54 billion and $1.89 billion respectively. This accounts for about 53% of the total ICT export amount of $14.61 billion. It means that if these two industries fall behind in global competition, the entire national ICT industry could be shaken.


The problem is that with the intensification of global competition, the fierce pursuit by foreign countries continues. In particular, China is the most threatening country. After declaring the rise of their semiconductor and display industries, they are quickly catching up with Korean companies. Indeed, market research firm 'UBI research' expressed concerns that by 2025, the shipment of domestic smartphone OLED panels might be surpassed by China. The threat from the United States, backed by enormous capital and national power, is also significant.


Therefore, to maintain its lead in the semiconductor and display industries, South Korea must continuously conduct advanced process technology research. Experts particularly emphasize that securing process technology based on 'Artificial Intelligence (AI)' will be the key to dominating the future semiconductor and display industries.

29일 서울대학교에서 열린 ‘제6회 반도체·디스플레이 공정진단제어 기술 연구회’ 현장./ 박설민 기자

◇ Complex Plasma Process, Efficiency Maximized with AI… Already Actively Used in the USA


At the '6th Semiconductor and Display Process Diagnosis and Control Technology Workshop' held at Seoul National University on the 29th, experts agreed that "AI will play an important role in plasma processes for making ultra-high-performance semiconductors and displays."


Plasma is a state where gas is heated to extremely high temperatures, separating into electrons and positively charged ions. It is also known as the 'fourth state of matter' and is unique in that it can be controlled at the atomic level using electromagnetic fields. This is a critical condition for manufacturing circuits in semiconductor and display processes, which are much finer than a strand of hair.


However, this makes it difficult to precisely control the plasma during the process. It involves depositing and etching layers on a substrate at the level of individual atoms. In such cases, it can be extremely difficult for engineers to discern defects with the naked eye.


The reason why 'plasma etching', which carves semiconductor substrates at the nanometer (nm) scale to create ultra-fine circuits, is considered one of the most challenging processes is precisely due to these factors. Experts emphasize the need to maximize the performance of semiconductor process diagnostics by utilizing AI's ability to analyze image data for this reason.

학회 좌장을 맡은 김곤호 반도체디스플레이 공정진단제어 기술연구회장./ 박설민 기자

Gonho Kim, chairman of the Semiconductor and Display Process Diagnosis and Control Technology Workshop and the session chair, said, "AI-based plasma process diagnostics are becoming a key factor in enhancing the competitiveness of the domestic semiconductor and display industry." He added, "This technology can give wings to the capabilities of outstanding domestic semiconductor process engineers."


In fact, related technology research is already actively underway overseas. The United States is currently leading in AI-based plasma process technology research. The 'Intelligent Equipment Research Team' at 'Lam Research', one of the world's top four semiconductor equipment manufacturers, is achieving tangible results.


The technology developed by the team led by Principal Researcher Andrew D. Bailey is for 'monitoring the Plasma-Enhanced Chemical Vapor Deposition (PECVD) process'. This technology involves using AI to analyze defects occurring in the PECVD process. It utilizes a machine learning model called the 'Hidden Markov Model (HMM)' to express changes in phenomena probabilistically. Simply put, it's a technology that analyzes the causes of defects when they occur and finds related data. According to Lam Research, the accuracy of defect detection is about 81%, which is considered very high considering that the accuracy of general engineers in defect analysis is around 60%.


Chairman Gonho Kim emphasized, "Although our country makes semiconductors and displays the best in the world, in order to continuously maintain our competitive edge overseas, we need to do what we are good at even better." He stressed, "The application of AI, a symbol of the Fourth Industrial Revolution, in the field of plasma process diagnostics is more necessary than ever."

머신러닝을 융합한 TCAD에 대해 발표하는 권형철 SK하이닉스 팀장./ 박설민 기자

Leading companies such as Samsung and SK Hynix recognize the importance of AI in the semiconductor and display industry, and national-level research focused on government-funded research institutes is also underway.


In response to this global competition, Korea's semiconductor and display industries are actively engaged in research. Samsung Display is at the forefront in this area and is currently adopting a technology called 'Plasma Information-based Virtual Measurement (PI-VM)' in their display process. This involves applying an AI model based on plasma physical information factors to actual mass production processes, and it is known to have solved specific process defects that occur during the process.


SK Hynix also recognizes the importance of AI-based semiconductor process diagnostics. According to Kwon Hyungcheol Kwon, team leader at SK Hynix, who presented at the conference, the current direction of their research is predicting memory cell characteristics using AI-based simulations. This simulation combines 'Technology Computer-Aided Design (TCAD)' with machine learning. TCAD refers to computer simulations conducted before semiconductor process technology or device development.


Team Leader Hyungcheol Kwon stated, "As the complexity of DRAM and NAND process and defect diagnostics increases, there is a growing need for technologies that utilize TCAD from the early stages of development for optimal condition searching and defect analysis." He added, "By applying deep learning AI models for image analysis to TCAD and running simulations on hundreds to thousands of cases, we have successfully achieved effective process diagnostics."


EDCORE, a company specializing in intelligent smart factory solutions, is also actively researching related technologies. In 2021, they launched the 'Intelligent Real-time Predictive System (Snotra VM)'. This system, based on AI reinforcement learning, features advanced multivariate process control algorithms, enabling autonomous monitoring within semiconductor manufacturing sites and the derivation of defect causes. According to EDCORE, the introduction of this system has improved process capability by more than 20% and quality by more than 30%, while reducing scrap rates by over 90%.

Deasoo Han, the representative, said, "To increase the Overall Equipment Effectiveness (OEE) in the semiconductor and display industry, the era has arrived where it is difficult to respond with the traditional manual-dependent systems." He emphasized, "In this field, achieving both productivity and quality while handling more and complex tasks will be made possible by AI technology."

발표를 진행하는 윤정식 플라즈마 공정장비 지능화 융합연구단장./ 박설민 기자

National research institutions are also actively conducting related research. A leading research group in this field is the 'Plasma Process Equipment Intelligence Convergence Research Group.' This group is led by the Korea Institute of Fusion Energy (KFE) and includes four government-funded research institutes - Korea Institute of Science and Technology (KIST), Korea Institute of Science and Technology Information (KISTI), and seven major universities. It was established with the support of the National Research Council of Science & Technology's convergence research group project, aiming to develop and demonstrate AI-based plasma process diagnostic equipment. The research group has a budget of 44 billion won allocated for research until 2026.

One of the representative technologies being developed by the research group is the 'AI-OES Sensor.' This system uses AI to monitor plasma changes within semiconductor and display process equipment.

The principle of this technology is similar to that of a smartwatch used for blood pressure measurement. Since the sensor of a smartwatch measures blood pressure on the surface of the skin, the readings are inevitably lower than the actual values. In this case, AI corrects these readings to derive values that are nearly 90% similar to the actual blood pressure. Similarly, the AI-OES sensor first observes the semiconductor and display process conditions within the process equipment in real time. Then, the AI applied to the sensor corrects the observed data, and provides engineers with highly accurate results.

Jungsik Yoon, head of the Plasma Process Equipment Intelligence Convergence Research Group, said, "In the global semiconductor and display competition, manufacturing and design technology are important, but ultimately, the leadership goes to the countries backed by mass production capabilities." He added, "To outpace China's aggressive pursuit, it will be necessary to introduce AI-based plasma process technology in the semiconductor and display fields."



bottom of page