Chinese semiconductor thread II
- Thread starter vincent
- Start date
Fast and Accurate EUVL Thick-Mask Model Based on Multi Channel Attention Network.
Abstract:
Simulation of thick-mask effects is an important task in computational lithography within extreme ultraviolet (EUV) waveband. This paper proposes a fast and accurate learning-based thick-mask model dubbed multi-channel block attention network (MCBA-Net) to solve this problem for EUV lithography. The proposed MCBA-Net introduces geometric feature attention module and structural feature attention module to improve the computation accuracy of thick-mask diffraction near field. During the training process, the proposed attention modules can effectively learn the impact of the three-dimensional mask diffraction behavior. In addition, the multi-channel network architecture is used to simultaneously synthesize the thick-mask diffraction matrices under different polarization states, and the coupling between different diffraction matrices is addressed. Numerical experiments show that the proposed model improves the computational efficiency by more than 20-fold over the rigorous simulator, and reduces the prediction error by 25% 50% compared with the state-of-the-art deep learning models. In addition, the generalization ability of the proposed method is proved using a complex testing pattern.
Machine Learning use in semiconductor manufacturing and design process.
With the continuous miniaturization of integrated circuit feature sizes, the computational complexity of process and device simulation has increased significantly. Traditional physical modeling methods are difficult to meet the needs of large-scale simulation and rapid iteration. It is urgent to introduce cutting-edge artificial intelligence technology and develop efficient simulation models to support advanced process parameter optimization and new device design.
In terms of simulation acceleration of etching process, Associate Researchers of the EDA Center conducted joint research and proposed a cascade recursive neural network (CRNN). This model fully characterizes the etching profile through a recursive neural network and realizes the information fusion of profile features and process parameters (such as pressure, power, temperature and voltage) through cascade combination layers. Through simulation and experimental data verification, compared with the simulation results of traditional etching models, the CRNN model significantly improves the modeling efficiency while ensuring the prediction accuracy, verifying the computational efficiency advantage of artificial intelligence in process modeling.
In terms of GAAFET device simulation acceleration, the team led by Associate Researcher from the EDA Center, in collaboration with the Beijing University of Aeronautics and Astronautics and ProLiant Electronics, proposed a densely connected deep neural network (DenseDNN), which provides stronger feature extraction and information transfer capabilities by establishing direct connections between the layers of the neural network. Combined with cost-sensitive learning methods, the model pays more attention to key data that determines device characteristics, thereby achieving accurate prediction of key device characteristics. This research work demonstrates the performance and efficiency advantages of artificial intelligence in device modeling, and provides a reference for intelligent models of advanced semiconductor devices.
The research results of intelligent driven etching process simulation, "Etching Process Prediction based on Cascade Recurrent Neural Network", were published in the authoritative journal of artificial intelligence, Engineering Applications of Artificial Intelligence (EAAI). The research results of intelligent device modeling, " Device Modeling Based on Cost-Sensitive Densely Connected Deep Neural Networks ", were published in the authoritative journal in the field of devices, IEEE Journal of the Electron Devices Society. Tang Xiaoying, a graduate student at the Institute of Microelectronics, is the first author, and Yao Zhenjie, an associate researcher at the Institute of Microelectronics, is the corresponding author. Both studies were supported by the National Key R&D Program and the Strategic A-level Priority Research Program of the Chinese Academy of Sciences.
Boline Narun announced that the "annual production of 18,000 tons of nano-silicon oxide, 34,000 tons of semiconductor CMP polishing liquid and 1.15 million square meters of CMP polishing pad materials project" has completed the topping-out of the first phase of the project.
Chemical Mechanical Polishing (CMP) technology is known as the only technology that can achieve global flattening of wafer surfaces in integrated circuit (IC) manufacturing in today's era. The effect of chemical mechanical polishing directly affects the final quality and yield of the chip. CMP removes surface defects at the atomic level and obtains a globally flattened surface through the dual coupling of chemical corrosion of chemical reagents in the polishing liquid and mechanical grinding . The polishing pad plays a role in storing polishing liquid, conveying polishing liquid, discharging waste, transferring processing loads, and ensuring the smooth progress of the polishing process during the CMP process.
Zhejiang Bolai Narun Electronic Materials Co., Ltd. focuses on providing overall solutions for planarization materials in the pan-semiconductor industry, and is committed to the technical research and development and industrialization of electronic-grade nano-silicon oxide abrasives, CMP polishing liquids for pan-semiconductors, polishing pads and other products , providing overall process material solutions for nano-level planarization of semiconductor substrates and other materials. The company has more than 10 years of CMP research and development and industrialization foundation. Its main products include silica sol, polishing liquid, and polishing pads . Its application areas include silicon carbide substrate CMP process, silicon substrate CMP process, integrated circuit CMP process, and other pan-semiconductor material CMP process.
Capital increase 1999 times! Beijing continues to focus on integrated circuits!
In terms of Beijing state-owned assets, Yizhuang Technology and Beijing State-owned Assets Management have become the second and third largest shareholders respectively. Beijing Economic and Technological Development Zone has become the highland of integrated circuit industry in southern Beijing, gathering nearly 400 integrated circuit companies, and the industry scale will exceed 100 billion yuan in 2024. Investing in Nortel Integrated will help promote the development of special processes and breakthroughs in key technologies. It is expected that the localization rate of China's integrated circuit market will continue to rise, and the gap between supply and demand of mature processes will increase. Nortel Integrated project products have broad application scenarios.
"It is estimated that by 2026, the localization rate of my country's integrated circuit market will rise to 21.2%. It is estimated that by 2027, the supply and demand gap of my country's 28nm and above mature processes will increase to about 370,000 pieces per month, and the market space is relatively large. Nortel's integrated project has clear product planning and target customers, and has broad application scenarios in the display driver IC, digital-analog hybrid chip, MCU and other markets." A relevant person revealed.
Nortel Integration is located in Beijing Economic and Technological Development Zone.
The district has quietly become an integrated circuit industry highland in southern Beijing. Currently, the district has gathered nearly 400 integrated circuit companies, forming an integrated circuit industry chain led by SMIC, North Huachuang, and Changxin Integrated Circuit, covering design, manufacturing, packaging and testing, equipment and parts, etc. In 2024, the scale of the integrated circuit industry in the Economic Development Zone will exceed 100 billion yuan for the first time, the industrial output value will exceed 90 billion yuan, and the industry growth rate will exceed 43%, ranking first in the country.
In terms of Beijing state-owned assets, Yizhuang Technology and Beijing State-owned Assets Management have become the second and third largest shareholders respectively. Beijing Economic and Technological Development Zone has become the highland of integrated circuit industry in southern Beijing, gathering nearly 400 integrated circuit companies, and the industry scale will exceed 100 billion yuan in 2024. Investing in Nortel Integrated will help promote the development of special processes and breakthroughs in key technologies. It is expected that the localization rate of China's integrated circuit market will continue to rise, and the gap between supply and demand of mature processes will increase. Nortel Integrated project products have broad application scenarios.
"It is estimated that by 2026, the localization rate of my country's integrated circuit market will rise to 21.2%. It is estimated that by 2027, the supply and demand gap of my country's 28nm and above mature processes will increase to about 370,000 pieces per month, and the market space is relatively large. Nortel's integrated project has clear product planning and target customers, and has broad application scenarios in the display driver IC, digital-analog hybrid chip, MCU and other markets." A relevant person revealed.
Nortel Integration is located in Beijing Economic and Technological Development Zone.
The district has quietly become an integrated circuit industry highland in southern Beijing. Currently, the district has gathered nearly 400 integrated circuit companies, forming an integrated circuit industry chain led by SMIC, North Huachuang, and Changxin Integrated Circuit, covering design, manufacturing, packaging and testing, equipment and parts, etc. In 2024, the scale of the integrated circuit industry in the Economic Development Zone will exceed 100 billion yuan for the first time, the industrial output value will exceed 90 billion yuan, and the industry growth rate will exceed 43%, ranking first in the country.