Abstract Chemical mechanical polishing (CMP) is critical for back-end-of-line (BEOL) integration, enabling planarization of dissimilar materials such as copper, tantalum nitride, TEOS SiO 2 , and the low- k dielectric BD1. Achieving optimal material removal rate (MRR) and surface quality requires understanding how pad stiffness, substrate properties, and interfacial interactions govern polishing. CMP of these representative BEOL materials was systematically studied using a slurry containing silica and polymer nanoparticles and three polishing pads differing in stiffness and surface morphology: IC1000 (rigid), VP3100 (intermediate), and Fujibo H800 (soft). Friction forces were recorded and analyzed in time and frequency domains to assess correlation with MRR. While mean friction and coefficient of friction (COF) varied with substrate and downforce, time-averaged metrics alone did not reliably predict MRR across pad–substrate combinations. Frequency-domain analysis revealed that power in dominant peaks strongly correlated with MRR, capturing the intensity of stick–slip interactions and the effect of pad compliance. IC1000 consistently exhibited higher dominant-peak power, explaining its superior MRR despite lower COF, whereas VP3100 showed weaker spectral coherence and reduced removal despite higher mean friction. Dominant peaks in the friction spectrum quantitatively link stick–slip dynamics and pad compliance to polishing efficiency, providing a mechanistic framework for CMP optimization.
Bhonsle et al. (2025) studied this question.