Virtual metrology models for predicting avera PECVD oxide film thickne

Ferreira, Ariane; Roussy, Agnès; Kernaflen, Christelle; Gleispach, Dietmar; Hayderer, Günter; Gris, Herve; Besnard, Jérôme

doi:10.1109/asmc.2011.5898187

Cited by 7 publications

(9 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A typical approach of FDC analysis is to stack up hundreds of traces of Fig.1 together and then define control limits of normal behaviours for the process, based on techniques of partial least squares (PLS) or others [1][2][3][4]. For this case, due to the lack of repeated measurements of wafer runs and taking advantage of the cyclical behaviours of the temporal traces, a refined practice is to stack each of the 7-cycles of the single trace together, as shown in Fig.2.…”

Section: A Data Alignments Of It Solutionsmentioning

confidence: 99%

“…It severely affects analytical computations of the entry and exit radii, r 1 (τ) and r 2 (τ) in (3). For this study, it suffices to make the simplest assumption of a uniform silhouette just to demonstrate what can possibly be achieved in engineering analysis, without resort to massive crunching of black-box data.…”

Section: Comparisonsmentioning

confidence: 99%

“…• (3) crack the data-segments for correlations using an arsenal of statistical techniques [1][2][3][4]; and…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Case of small-data analysis for ion implanters in the era of big-data FDC

Hui

Mou

2013

ASMC 2013 SEMI Advanced Semiconductor Manufacturing Conference

View full text Add to dashboard Cite

This paper presents a case study of constructing process models based on physical mechanisms of semiconductor manufacturing tools in attempts to predict behaviours of process conditions. Actual measurements from the processing tools are always corrupted with noises and crunching huge volumes of temporal traces of status variables very often fail to pinpoint the accurate fault conditions, not to mention any of their efficient classifications, should abnormal conditions really exist. The current fashion of moving into massive big data computing is yet to distill concrete correlations among tool conditions and impacts on process results of semiconductor devices. As an alternative before the foolproof maturity of big data cracking, and in contrast to the conventional black-box approach of statistical regressions, we take a fundamental view in constructing physical model of the ion implantation process for a flywheel implanter, first to calculate the motion trajectories and subsequently, the implantation dosage on the wafer. We summarize the underlying solution techniques in principles and leave the specific details of parameter calibrations to individual field practitioners. Keywords -advanced process control, fault detection and classification, model-based controlI. 978-1-4673-5007-5/13/$31.00

show abstract

Section: A Data Alignments Of It Solutionsmentioning

confidence: 99%

Section: Comparisonsmentioning

confidence: 99%

See 1 more Smart Citation

Case of small-data analysis for ion implanters in the era of big-data FDC

Hui

Mou

2013

ASMC 2013 SEMI Advanced Semiconductor Manufacturing Conference

View full text Add to dashboard Cite

show abstract

“…Although this algorithm does not provide the best fitting accuracy, it does provide an interpretable model that can be beneficial to the engineer for gaining an easy-to-understand processing knowledge. The most often studied linear regression algorithms include Multiple Linear Regression 23,25 (MLR), Least Absolute Shrinkage and Selection Operator (LASSO) regression [26][27][28][29] , Partial Linear Square Regression (PLS) 23,23,[29][30][31][32][33][34] , Support Vector Regression (SVR) 23,29,35,36 and Ridge Linear (RL) regression 23 , etc.…”

Section: Introductionmentioning

confidence: 99%

“…Decision tree regression provides a good balance between interpretability and fitting accuracy comparing to linear and network based regressions. It includes standard decision trees regression 26,36 , Random Forest regression 26 , Gradient Boosting (GB) regression 37 and Tree Ensemble regression 32,33,38 , etc.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning based CVD Virtual Metrology in Mass Produced Semiconductor Process

Xie,

Stearrett

2021

Preprint

View full text Add to dashboard Cite

A cross-benchmark has been done on three critical aspects, data imputing, feature selection and regression algorithms, for machine learning based chemical vapor deposition (CVD) virtual metrology (VM). The result reveals that linear feature selection regression algorithm would extensively under-fit the VM data. Data imputing is also necessary to achieve a higher prediction accuracy as the data availability is only ∼70% when optimal accuracy is obtained. This work suggests a nonlinear feature selection and regression algorithm combined with nearest data imputing algorithm would provide a prediction accuracy as high as 0.7. This would lead to 70% reduced CVD processing variation, which is believed to will lead to reduced frequency of physical metrology as well as more reliable mass-produced wafer with improved quality.

show abstract