A stack fusion model for material removal rate prediction in chemical-mechanical planarization process

Zhao, Shuai; Huang, Yongfeng

doi:10.1007/s00170-018-2578-5

Cited by 10 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Assuming that the parameters of Σ X,X are θ, and Σ X,X is replaced by Σ θ , while the input data are considered to have completed normalization with a mean of 0, according to Equations (3)- (5), y conforms to a Gaussian distribution with a mean of 0 and a variance of Σ θ + σ 2 I. The probability distribution function of y is shown in Equation (6):…”

Section: Related Work 21 Deep Kernel Learningmentioning

confidence: 99%

“…Since individual machine learning models do not suitably fit the CMP process, some studies have used stacking models. Zhao and Huang performed one-hot encoding of the "stage" and "chamber" variables in the feature creation and feature encoding stages to transform the process data into multidimensional information, and the stacking integration model was chosen for the regression [5]. Li and Wu et al also used a stacking integrated learning regression model with primary learners including random forest (RF), gradient boosting tree (GBT), and extreme random tree (ERT), and meta-learner as extreme learning machine (ELM) and classification and regression tree (CART), and the features in the model included the frequency domain features of the three rotation variables and feature selection using RF [6].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Semi-Supervised Deep Kernel Active Learning for Material Removal Rate Prediction in Chemical Mechanical Planarization

Huang

Zhang

et al. 2023

Sensors

View full text Add to dashboard Cite

The material removal rate (MRR) is an important variable but difficult to measure in the chemical–mechanical planarization (CMP) process. Most data-based virtual metrology (VM) methods ignore the large number of unlabeled samples, resulting in a waste of information. In this paper, the semi-supervised deep kernel active learning (SSDKAL) model is proposed. Clustering-based phase partition and phase-matching algorithms are used for the initial feature extraction, and a deep network is used to replace the kernel of Gaussian process regression so as to extract hidden deep features. Semi-supervised regression and active learning sample selection strategies are applied to make full use of information on the unlabeled samples. The experimental results of the CMP process dataset validate the effectiveness of the proposed method. Compared with supervised regression and co-training-based semi-supervised regression algorithms, the proposed model has a lower mean square error with different labeled sample proportions. Compared with other frameworks proposed in the literature, such as physics-based VM models, Gaussian-process-based regression models, and stacking models, the proposed method achieves better prediction results without using all the labeled samples.

show abstract

Section: Related Work 21 Deep Kernel Learningmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Semi-Supervised Deep Kernel Active Learning for Material Removal Rate Prediction in Chemical Mechanical Planarization

Huang

Zhang

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Predicting the removal rate and friction coefficient is essential for monitoring and automating the polishing process. Therefore, previous studies have focused on methods for predicting the removal rate based on neural networks, 23 the wafer-scale multi-physics model, 24 the stack fusion model, 25 and the reference-based virtual metrology method. 26 Other studies have considered the singlecorrelation between the friction coefficient and platen load current associated with the CMP of copper, cobalt, and STI (shallow trench isolation) 18 and CMP of tungsten and interlayer dielectrics.…”

mentioning

confidence: 99%

Real-Time Prediction of Removal Rate and Friction Coefficient During Chemical Mechanical Polishing Using Motor Load Currents with a Polisher

Uneda¹,

Ota²,

Takiguchi³

et al. 2023

ECS J. Solid State Sci. Technol.

View full text Add to dashboard Cite

Herein, a method for predicting real-time removal rate and friction coefficient between the pad and substrate during chemical mechanical polishing was investigated using only the load currents of two motors of a polisher. Polishers for semiconductor devices are equipped with various sensors, enabling a real-time prediction of the removal amount. The polishers used to polish substrates are not usually equipped with sensors, and the polishing time is fine-tuned by skilled-technicians to achieve the desired substrate thickness. However, since every polisher has some motors, predicting the removal rate and friction coefficient using only the real-time data produced by these motors would be beneficial. This study attempts to predict the removal rate and friction coefficient in long-time polishing using a training dataset obtained from short-time polishing. Results showed that by performing extremely low-pressure, long-time polishing to understand the polisher characteristics and then subtracting the polisher characteristics from the motor information during long-time polishing, highly accurate predictions of the removal rate and friction coefficient within ~94% in percent match (prediction accuracy) between the experimental and predicted values can be obtained. Furthermore, slurry degradation during CMP can be monitored using this prediction method.

show abstract

Phase Partition Based Virtual Metrology for Material Removal Rate Prediction in Chemical Mechanical Planarization Process

Jiang

Zhang

et al. 2021

Artificial Intelligence

View full text Add to dashboard Cite

A stack fusion model for material removal rate prediction in chemical-mechanical planarization process

Cited by 10 publications

References 9 publications

Semi-Supervised Deep Kernel Active Learning for Material Removal Rate Prediction in Chemical Mechanical Planarization

Semi-Supervised Deep Kernel Active Learning for Material Removal Rate Prediction in Chemical Mechanical Planarization

Real-Time Prediction of Removal Rate and Friction Coefficient During Chemical Mechanical Polishing Using Motor Load Currents with a Polisher

Phase Partition Based Virtual Metrology for Material Removal Rate Prediction in Chemical Mechanical Planarization Process

Contact Info

Product

Resources

About