Controlling Galvanic Corrosion in Post-Copper-CMP Cleaning
Maria A. Lester, Associate Editor -- Semiconductor International, 11/1/2000
 |
It occurred only in the device wafer, not the structural test wafer. The researchers observed a new type of corrosion mechanism during post-CMP cleaning, and they found that this corrosion was induced by the heating lamp during the spin-rinse-dry process. They removed the corrosion byproduct with a solvent clean.
They found and prevented light-induced corrosion through several steps:
1. Slurry behavior. The wafer surface was free of corrosion after the copper removal. The researchers detected corrosion after the barrier metal removal, finding it to be independent of the slurry. They found that the byproduct was mainly copper. They concluded that this type of corrosion couldn't be prevented.
2. Cleaning chemicals behavior. The researchers tested three different chemicals for cleaning the wafers, but none could prevent corrosion. Therefore, they concluded that the corrosion is independent of the cleaning chemicals. However, they could remove the corrosion byproduct by solvent cleaning with an integrated scrubber-type post-CMP cleaner with the baseline chemical if the wafer was drying with only hot nitrogen instead of the spin-rinse-dry process. The researchers suspected that the spin-rinse-dry process was the cause of the byproduct.
3. Spin-rinse-dry process behavior. Wafers were corrosion-free after drying with hot nitrogen only. Corrosion was also prevented using hot nitrogen and the spin-dry process. This made the researchers suspect the heating lamp as the cause of corrosion.
4. Heating lamp behavior. The corroded wafer was processed through the cleaner again, resulting in larger corrosion byproduct. The light generated electron-hole pairs in the depletion region of the P/N junction, creating an active galvanic cell when the wafer was still wet. The researchers observed copper lines on the P+ and N+ regions behaving as anode and cathode, respectively. And they saw copper lines bridging after the removal of barrier metal. Electron-hole pairs could not be generated when the lamp was off and no corrosion was observed. Because the copper lines were electrically separated after the removal of the barrier, this type of corrosion occurred only in the device wafer. They found that the galvanic cell could be disconnected when the continuous wet path was removed.
Results indicate that the corrosion byproduct can be removed with solvent cleaning or by changing the process recipe in the spin-rinse-dry process. However, this new corrosion mechanism and the design rule of metal lines need further investigation.
