Computational Lithography Using Machine Learning Models

Shin, Youngsoo

doi:10.2197/ipsjtsldm.14.2

Cited by 8 publications

(3 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent decades, machine learning (ML) and artificial intelligence (AI) have come up with impressive practical examples ranging from image recognition to strategy games and bio-medical applications. Many of industrial applications are based on image recognition/optimization algorithms, which for example is widely utilized in modern lithography starting from mask optimization with optical proximity correction (OPC) and etch proximity correction (EPC), assist features insertion and their printability check, hotspot detection 1) and down to TD-CDSEM (top-down critical dimension scanning electron microscope) image quality enhancement. 2,3) Even though it is less dependent on imaging, the field of plasma etching is not an exception that was put off the game.…”

Section: Introductionmentioning

confidence: 99%

Assessment of STI dry etch process variability by means of dynamic time warping technique

Milenin¹,

Chan²,

Lazzarino³

2023

Jpn. J. Appl. Phys.

View full text Add to dashboard Cite

Numerical data of two in-situ optical acquisition systems were used in machine learning algorithm to evaluate an shallow trench isolation dry etch process in a dataset of more than 200 etched wafers processed during a year. Though overall recipe performance was according to specifications, the observed parameter fluctuations were characterized to check for correlation patterns in processing data by means of machine learning. Thus, dynamic time wrapping was used to analyze time series datasets to get characteristics of minimum distance’s path between signals coming of each individual wafer. Such metrics of distances could help clustering groups of wafers that appear to be in proximity. Out of eight analyzed channels, we observed the largest distance variability was in the Si main etch for spectral reflectometer dataset, which we compared to optical emission spectroscopy channels. Based on correlation assessment of distances to the etch depth uniformity data, we showed that the selected approach is a good means to analyze time series datasets of e.g. dry etch processes for monitoring process stability.

show abstract

Section: Introductionmentioning

confidence: 99%

Assessment of STI dry etch process variability by means of dynamic time warping technique

Milenin¹,

Chan²,

Lazzarino³

2023

Jpn. J. Appl. Phys.

View full text Add to dashboard Cite

show abstract

“…Another point that should be mentioned is that the main analysis process in these strategies is based on an artificial neural network, or a convolutional neural network. 26,29) The prediction accuracy is high. However, the physical meaning is unclear.…”

Section: Introductionmentioning

confidence: 99%

Classification of lines, spaces, and edges of resist patterns in scanning electron microscopy images using unsupervised machine learning

Jin

Kozawa

2022

Jpn. J. Appl. Phys.

View full text Add to dashboard Cite

As key steps of lithography, the development of resist materials and the exploration of new materials are important to meet market demands from semiconductor industry. During the development, resist materials are usually evaluated by the information extracted from their scanning electron microscopy (SEM) images. The information extracted from SEM images is not always accurate owing to the technical limitation. Accurate information extraction is also useful for the prediction of an etched substrate pattern. In this paper, we reported a strategy to classify the image pixels of line-and-space resist patterns into line, space, and edge classes, using unsupervised machine learning. Brightness and coordination information was integrated into the classification method. The high reliability in classification was demonstrated by hierarchical clustering based on its information integrating ability. Among all the methods of hierarchical clustering examined, the centroid method was the most accurate strategy for extracting information from a single SEM image.

show abstract

“…Other striking examples come from the field of computational lithography, where the Volterra-Wiener type of models [20]- [22] and convolutional neural networks [23]- [28] have been employed to simulate the diffusive and nonlinear process of photochemistry, which turns an optical image of photoexposure in a photoresist material to a 3D topography of a developed photoresist. There, the optical image is sampled at near or slightly over the Nyquist rate determined by a bandwidth limit of the optical imaging system, with pixels sized on the order of tens of nanometers, while the process of photochemistry is highly nonlinear to induce seemingly unavoidable aliasing, such that a miniscule coordinate shift by a small fraction of the pixel size could induce a significant change in the results, manifesting themselves in lithographic patterns having size or position errors approaching even exceeding a nanometer, becoming unacceptable in advanced high-density lithography processes.…”

mentioning

confidence: 99%

Aliasing-Free Nonlinear Signal Processing Using Implicitly Defined Functions

Wei

2022

IEEE Access

View full text Add to dashboard Cite

Digital signal processing relies on the Nyquist-Shannon sampling theorem that applies to and requires a continuous signal with a limited bandwidth. However, many systems or networks of signal processing involve nonlinear functions, which could generate new frequency components beyond the original bandwidth and lead to aliasing. Indeed, aliasing-induced shift variance has long been a nuisance and unsolved problem in convolutional neural networks, and recently been found to severely impair the performance of machine learning applications. The same problem exists in other fields such as computational lithography. In this paper, a new method and algorithms are introduced to solve the problem of aliasing induced by nonlinear functions involving operations other than linear convolutions and pointwise multiplications. Said new method and algorithms employ implicitly defined functions that are implemented via iterations of polynomial operations, so that aliasing is completely avoided by upsampling signals before polynomial operations and limiting signal spectra before downsampling. Theoretical analyses and exemplary algorithms are presented to implement nonlinear functions commonly used in signal processing networks. In particular, exemplary embodiments and numerical experiments are reported to illustrate and verify aliasing-free operations of Wiener-Padé approximants, which are already universal in their ability to approximate any continuous activation functions to a desired accuracy.

show abstract

Computational Lithography Using Machine Learning Models

Cited by 8 publications

References 34 publications

Assessment of STI dry etch process variability by means of dynamic time warping technique

Assessment of STI dry etch process variability by means of dynamic time warping technique

Classification of lines, spaces, and edges of resist patterns in scanning electron microscopy images using unsupervised machine learning

Aliasing-Free Nonlinear Signal Processing Using Implicitly Defined Functions

Contact Info

Product

Resources

About