EUVIMEDEUVIMED entdecken
Crossref

Computational and Experimental Study of Velocity Anisotropy and Slurry Hydrodynamics in Wafer-Pad CMP: Rs-Governed Material Removal with Machine Learning Prediction

Jordan Journal of Mechanical and Industrial Engineering · 05.03.2026 · 17

DOI: 10.59038/jjmie/200102 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Jordan Journal of Mechanical and Industrial Engineering, online erschienen am 05.03.2026.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Chemical Mechanical Polishing (CMP) remains a critical challenge in advanced semiconductor manufacturing, particularly for leading chip producers such as NVIDIA and TSMC, where sub-5 nm architectures, HBM stacking, and chiplet integration demand exceptionally tight planarity control. The failure of traditional CMP models based on Preston's equation to account for nonlinear slurry–pad–wafer interactions, velocity anisotropy, and pad wear variability results in defects such dishing, erosion, and non-uniform thinning. To get beyond these restrictions, this work offers a unique kinematic–computational–machine learning (ML) framework. Analysis of the slurry-flow behavior and velocity distribution over a 100 mm wafer at various RPM ratios shows that Rs = 1.0 minimizes velocity non-uniformity, whereas Rs > 1 increases MRR and Rs < 1 decreases it. High accuracy MRR prediction is produced by a Random Forest model with LOOCV, and the Pad–Wafer RPM Ratio is found to be the primary optimization variable. Additionally, a new sweeping-rotation technique is presented to maintain planarization and enhance shear distribution. The suggested methodology directly addresses the CMP limitations encountered by next-generation GPU and semiconductor production by enabling AI-driven CMP optimization, which offers greater uniformity, decreased defectivity, and increased yield.

Zitieren

APA

(2026). Computational and Experimental Study of Velocity Anisotropy and Slurry Hydrodynamics in Wafer-Pad CMP: Rs-Governed Material Removal with Machine Learning Prediction. Jordan Journal of Mechanical and Industrial Engineering, 17. https://doi.org/10.59038/jjmie/200102

Vancouver

Computational and Experimental Study of Velocity Anisotropy and Slurry Hydrodynamics in Wafer-Pad CMP: Rs-Governed Material Removal with Machine Learning Prediction. Jordan Journal of Mechanical and Industrial Engineering. 2026:17. doi:10.59038/jjmie/200102

BibTeX
@article{euvimed2026,
  title = {Computational and Experimental Study of Velocity Anisotropy and Slurry Hydrodynamics in Wafer-Pad CMP: Rs-Governed Material Removal with Machine Learning Prediction},
  journal = {Jordan Journal of Mechanical and Industrial Engineering},
  year = {2026},
  pages = {17},
  doi = {10.59038/jjmie/200102}
}
Kostenlos starten →