1. Material Selectivity Control: Constructing an "Intelligent Polishing Barrier"
•Dual-Component Slurry Design: A synergistic system using phosphoric acid-based compounds (e.g., ammonium phosphate) and linear anionic polymers (e.g., polyacrylic acid) is employed. By electrostatically adsorbing onto the silicon nitride layer, it forms a protective film to suppress excessive polishing. Experimental data shows this scheme stabilizes the silica/silicon nitride polishing rate ratio between 30:1 and 50:1, preventing damage to underlying structures.
•Nanoscale Particle Modification Technology: Al₂O₃ coatings (5-10 nm thick) are applied to SiO₂ abrasives, enhancing hardness while reducing grinding fragmentation and edge chipping risk (validated in SMIC’s STI process).
2. Dynamic Process Parameter Control: Building an "Adaptive Polishing Engine"
•Pressure Gradient Distribution: Multi-zone backpressure control technology applies 5% additional pressure to the wafer edge to compensate for centrifugal force, while the central region maintains baseline pressure (1-3 psi for copper, 3-5 psi for silica), effectively suppressing edge over-polishing.
•Intelligent Rotation Ratio Adjustment: The rotation ratio is dynamically tuned (1:1 to 2:1) based on real-time thickness feedback. A 2:1 high-speed ratio removes protrusions in the initial stage, while a 1:1 low-speed ratio enables fine finishing in the final stage, balancing removal rate and precision.
3. Real-Time Monitoring and Closed-Loop Feedback: Developing a "Nanoscale Early Warning System"
•Multi-Modal Sensing Array: Integrates laser interferometry (detecting thickness changes ±0.5 nm), in-situ atomic force microscopy (AFM) scanning (sampling rate 100 μm²/s), and ellipsometry (refractive index monitoring) to reconstruct 3D topography in real time.
•Deep Reinforcement Learning (DRL) Control: A DRL model trained on historical data predicts over-polishing risks and automatically adjusts parameters. A TSMC case study shows this reduces dishing in 3nm-node Cu interconnect layers to <10 nm.
4. Equipment Structural Innovation: Developing an "Adaptive Polishing Interface"
•Piezoelectric Actuated Polishing Pad: Integrates micro-scale piezoelectric actuator arrays (<100 μm) to modify local topography via voltage control, compensating for wafer surface undulations.
•Dynamic Fluid Distribution System: Uses a 256-channel micro-nozzle array to adjust slurry flow rate (±5% precision) based on local pressure distribution, ensuring uniform reactant concentration.
5. Precise Chemical Environment Control: Establishing a "Stable Reaction Chamber"
•Dynamic pH Buffering System: Employs phosphate + borate composite buffers to limit pH fluctuations within ±0.1, suppressing metal ion-catalyzed reactions (e.g., Cu²⁺ concentration <1 ppb).
•Oxidizer Stabilization Technology: Adds stannate stabilizers to extend H₂O₂ half-life to 6 months, with a decomposition rate <3%/day, ensuring reaction controllability.
6. Data-Driven Process Optimization: Building a "Digital Twin System"
•Virtual Mapping Technology: Develops multi-physics simulation models of the wafer-polishing pad-equipment system to predict material removal distribution under varying parameters, reducing process development cycles by 40%.
•Defect Prediction Algorithm: Uses convolutional neural networks (CNN) to analyze AFM data, providing early warnings of over-polishing risks 30 minutes in advance with >95% accuracy.