Multivariable Process Monitoring to Enhance CMP Yield
Successful fabrication of high-performance, complex semiconductor devices relies on chemical mechanical planarization (CMP) to planarize the wafer surface before depositing the next layer of metal or dielectric. Trends in lithography are putting more pressure on CMP process steps. Decreases in line width and spacing of metal features push the limits of 193-nm immersion lithography. As a result, wafers are undergoing multiple CMP steps for each layer.
While CMP is necessary, it can also contribute to yield loss if the process is not tightly controlled and monitored. As the number of CMP steps increases and the thickness of material being removed decreases, CMP defect reduction becomes a greater challenge. The more complex the chip architectures and the smaller the critical feature size, the greater the risk. With less room for error, process optimization of CMP slurries is necessary to achieve desired yields.
Tight control over slurry properties including chemistry, working particle size, and particle size distribution helps achieve the desired removal rate and minimize CMP-induced defects. For leading-edge devices, the working particle size in a CMP slurry typically ranges from 30 to 200 nm, which means that even sub-micron particles are considered large from the defect risk viewpoint. Because CMP removes extremely thin layers of material, slurry stability is crucial for maintaining an appropriate removal rate.
The balance of process speed, performance, and cost tilts toward performance when particles around 100 nm in diameter can cause killer defects. Precise control of slurry chemistry and working particle size is the key to enhancing yield. Slurries must have a narrow, uniform particle size distribution and a reliable, consistent balance of chemical components.
This white paper discusses three types of process data analytics that can help ensure high CMP yields:
- Electrochemical analysis during blending
- In-line process chemistry monitoring
- Particle sizing determination