Detecting pad-wafer contact in CMP using dual emission laser induced fluorescence

Chemical Mechanical Polishing (CMP) is a technique used for planarizing a wide range of surfaces, including metal and dielectric materials, during the manufacturing of integrated circuits (ICs). Material removal in CMP occurs with the combind chemical and mechanical action of the slurry and polishing pad. For inter-layer dielectric (ILD) polishing, particle-wafer contact is believed responsible for planarization. The particles sit atop the polishing pad asperities, meaning removal rate should correlate with pad-wafer contact area. In this thesis, we presents in-situ measurements of slurry layer thickness and pad-wafer contact using Dual Emission Laser Induced Fluorescence (DELIF).; The DELIF technique for measuring the thickness of thin fluid films has been adapted to make instantaneous measurements of slurry thickness in CMP at high spatial resolution (2.9 mum per pixel). Modeling and calibration experiments confirm that there is a linear relationship between CMP DELIF image intensity and fluid layer thickness over a range of 0 to 133 mum. The slurry thickness in the contacting portions of a Cabot Microelectronics D100 polishing pad is 0-60 mum for polishing speeds ≤ 0.62 m/s and applied wafer load ≤ 1.7 psi. As the polishing pad becomes fully conditioned, the slurry layer thickness approaches a saturation point. The time to saturation is approximately 1 hour.; DELIF images of the slurry thickness can be analyzed for contact by plotting the image intensity histogram. This histogram is representative of the pad asperity height distribution. When the pad is in contact with the wafer, the asperity peaks flatten and give rise to a secondary peak in the low intensity extreme of the distribution representative of contact. Data simulations show that the secondary peak is sharp at low noise:signal ratio (<10% noise), and is gradually smoothed as more noise is added increasing measurement errors. Even though DELIF is a high noise measurement technique (≥ 15%), quantitative contact measurements can still be made if the measurement noise does not completely smooth the contact peak in the histogram. We found that the measurable contact area percent decreases with increasing conditioning time, increases with increasing pressure, and decreases with increasing velocity. The mean measured contact percentage for all tested process conditions was less than 1%. Due to high measurement noise associated with DELIF, contact below 0.1% cannot be measured.