Search

Post cleaning experiments for front end of the line (FEOL) CMP with silica and ceria slurries are carried out on commercial polishers with 300 mm oxide, nitride, and integrated shallow trench isolation (STI) wafers. Considering the charge attraction or repulsion between particles and wafer surface, both acidic and basic clean chemicals are applied at different stages in the post CMP process sequence. Diluted hydrogen peroxide in a non-contact megasonic cleaner is used to remove ceria abrasive particles and polish residues with high efficiency. On-platen buff clean with or without pad conditioning can make an impact on the post CMP cleaning performance. Post CMP cleaning splits are executed in order to evaluate the effectiveness of each of the process steps and their roles in the overall cleaning performance.

Poly silicon is the preferred gate material in terms of fabrication process in the fab, and gate height control is one of the factors determining transistor performance. If we can have a better uniformity control of poly CMP, it will be beneficial not only to CMP but also to the downstream etch process. Thus, we evaluated two poly slurries (Slurry A, B), analyzed each characteristic and performance on the device area and applied to 7nm CMP process furthermore. We found that slurry A, ceria-based poly slurry, has excellent WiD uniformity, less than 2nm, and good end-point (motor torque) detection point which is good in terms of manufacturing perspective. However, shelf life is a big challenge that we loss the selectivity of nitride and oxide after 15 days and also found visible white precipitate in the slurry loop. We did the shelf life test for both the additive only and the slurry. Data shows that the removal rate, pH and particle size of the additive only did not change a lot; however, the slurry became more unstable after 11 days with the increasing of the nitride loss, ten times the baseline. Chemical reaction happened between the abrasive and the additive. We can avoid the slurry deterioration by using the in-line mixing instead of pre-mixing. Table 2 shows the center die uniformity of slurry A is 1.9nm, much less than our target, 5nm. We also measured the CMP house pad topography for the slurry marginality test. Slurry A also has the better capability to stop on nitride, however, slurry B, the silica based slurry, has no removal rate at donut and center area. It can be caused by the slurry distribution on the pad or the high additive concentration. We will have a completed analysis of the ratio of slurry B and additive analysis and also dig more deeply about how slurry react with additive and what factor would be the major driving force of poly process in the full paper. Figure 1

The surface Cu flake particle generation mechanism during Cu Chemical Mechanical Planarization (CMP) process was studied to propose optimal tribological conditions. Cu flake defects are the main impact on yield and reliability in Logic Devices. Fig.1 shows typical Cu Flake images and their components. The particle components are copper and copper oxide based on EDX analysis with Cu flake sizes ranging from a hundred nm to a few um. Engineering polishing wafers with five kinds of pattern densities and three different line widths were prepared to evaluate defect density changes and elucidate the effects of surface topography and geometry. All polishing experiments were performed on a 300 mm polisher (LKPS, AMAT) with Hitachi Cu Barrier Slurry (HS-915TS). Defect density changes were calculated by two steps: 1) scanned with KLA-Tencor 2915 Bright Field Inspection Tool, 2) reviewed by Applied-Materials SEMVision with 30 nm horizontal resolution. Determining the defect generation mechanism was approached by three different methodologies. Firstly, wafer friction torque was analyzed to estimate the mechanical impact dependency of defect generation during the brushing part of the cleaning process with these different patterns. Usually, friction torque is a well-known method to check the macro-scale impact in the defect investigation process. However, there were no correlation between torque difference and defect density. It is assumed that the wafer level friction sensor could not detect the patterning pitch scale’s impact difference. Secondly, a nano-scale approaching method was utilized to investigate the Cu flake phenomena. Interestingly, the highest defect density showed in the smallest patterned density and in the 72 nm line width by repeatable tests. What it found was that the Cu protrusion amount, as shown in Fig.2, linearly decreases with pattern density, with 3 nm range of difference between the minimum and maximum pattern density. It could be suspected that a higher Cu protrusion profile may be one of the big distributors in making more mechanical impaction in sub 100 nm pitch scales. To verify protrusion volume effects, the slurry selectivity was changed to encourage more protrusion by changing the H2O2 mixing ratio. Experimental results showed that higher H2O2 resulted in a lower defect density. Lastly, the sliding velocity effect was studied with three different conditions: 1.0, 1.2, and 1.35 m/s. Fig.3 clearly showed the density effect of flake generation with each patterning wafer. In particular, a higher pattern density did not show defect density changes during the sliding speed tests, while a lower density pattern had increasing defects with higher sliding speeds. These results can explain the Cu flake generation mechanism by mechanical impact force, e.g. higher velocity speed and higher shear force during wafer cleaning process. We conclude that lower speed and less pattern selectivity is preferable to reduce the Cu flake defect level to overcome various pattern densities in the real production. We will explain in detail the experimental results and mechanism in the presentation. Figure 1