Process optimization of line patterns in extreme ultraviolet lithography using machine learning and a simulated annealing algorithm

Abstract

Resolution, line edge/width roughness, and sensitivity (RLS) are critical indicators for evaluating the imaging performance of resists. As the technology node gradually shrinks, stricter indicator control is required for high-resolution imaging. However, current research can improve only part of the RLS indicators of resists for line patterns, and it is difficult to improve the overall imaging performance of resists in extreme ultraviolet lithography. Here, we report a lithographic process optimization system of line patterns, where RLS models are first established by adopting a machine learning method, and then these models are optimized using a simulated annealing algorithm. Finally, the process parameter combination with optimal imaging quality of line patterns can be obtained. This system can control resist RLS indicators, and it exhibits high optimization accuracy, which facilitates the reduction of process optimization time and cost and accelerates the development of the lithography process.

Machine learning in electron beam lithography to boost photoresist formulation design for high-resolution patterning

“…At present, researchers have begun to use AI/ML to improve the quality of photoresist patterns, mainly adopting modeling and optimization methods based on multilayer feed-forward neural networks (MFNNs). 29,30,40,41 However, the accuracy of the MFNN model needs to be further improved.…”

Machine Learning Applied to Electron Beam Lithography to Accelerate Process Optimization of a Contact Hole Layer

“…Machine learning (ML) is one of the most intelligent and cutting-edge modeling methods in artificial intelligence (AI), which studies how to use computers to simulate or implement human learning activities . Recently, researchers have begun to pay attention to an ML method, that is, long short-term memory (LSTM) networks, and use them to build predictive models for fuel cells, financial market, stock market, soft sensor, wind power, photovoltaic power, and solar irradiance. − It is worth noting that many scholars have demonstrated that the LSTM network − with long-term memory function has better prediction ability than multilayer perceptron neural network (MPNN), ,,− which is one of the most commonly used ML modeling methods. Therefore, in this article, we use an LSTM network to establish a lithographic imaging prediction model.…”

“…Lithography is an imaging technology that transfers the designed circuit patterns to the wafer, which is considered to be the most critical process in chip manufacturing . At present, extreme ultraviolet (EUV) lithography and electron beam lithography (EBL) are regarded as two promising candidate technologies to realize high-resolution circuit patterns in a chip. − However, a large number of research results show that the imaging performance of the two lithography technologies is seriously affected by exposure, baking, development, and other process conditions. − As the feature size gradually decreases, the impact on lithography imaging will be more significant. Optimizing and controlling the lithographic process conditions have become an urgent need.…”

Critical dimension prediction of metal oxide nanoparticle photoresists for electron beam lithography using a recurrent neural network

“…Nowadays, machine learning methods are gradually being applied to the field of lithography, in which the most widely used is the multilayer feed-forward neural network (MFNN). At present, researchers have proposed process optimization approaches 5,21,[29][30][31][32][33] based on MFNN to obtain matching process conditions. Nevertheless, the MFNN model requires a large number of data sets to achieve high-precision predictive performance.…”