Paper PS-TuP21
Prediction of Particle Generation by Machine Learning in Plasma Etching Tools

Tuesday, October 31, 2017, 6:30 pm, Room Central Hall

Prognostics techniques, which predict the remaining useful life of the components and monitor the health conditions of process equipment by utilizing data-sets that acquired from the sensors of equipment components, are gaining attention to solve the cost issues in semiconductor manufacturing. [1].

In this study, development of a prognostics system to predict the health conditions that deviated over time in microwave plasma etching tools was investigated. The selection of the right analytical engines for getting effective results has been a major issue that impedes deployment of prognostics techniques. Several machine learning algorithms including PCA-based T-squared /square prediction error (SPE) [2], self-organizing map (SOM) - minimum quantization error (MQE) [3], auto-associative kernel regression (AAKR) [4] were evaluated to predict particle generation for the etching of metal layers such as work function metals (WFMs). The benchmarking results indicated that AAKR and PCA-T squared most effectively captured particle generation and showed better monitoring performance compared with other algorithms. In addition, process parameters that affect the particle generation were clarified by calculating contribution values for each process parameter. Details will be discussed in this presentation.

[1] Jay Lee et al., “Recent advances and trends in predictive manufacturing systems in big data environment,” Manufacturing Letters Volume 1, Issue 1, October 2013, Pages 38–41

[2] Kevin P. Murphy, “Machine Learning, A Probabilistic Perspective," The MIT Press, 2012)

[3] T. Kohonen, “The Self-Organizing Map,” Proceeding of the IEEE, vol. 78. pp. 1464–1480, 1990.

[4] P. Guo and N. Bai, “Wind turbine gearbox condition monitoring with AAKR and moving window statistic methods,” Energies, vol. 4, no. 11, pp. 2077–2093, 2011.