One major challenge in plasma etching processes for integrated circuit fabrication is to achieve a good wafer-to-wafer repeatability. This requires a perfect control of the plasma chamber wall conditions. For silicon etching processes, which deposit SiOyClz layers on the chamber walls, this is achieved by cleaning the interior surfaces of the plasma chamber with an SF6-based plasma after each wafer is etched. However, x-ray photoelectron spectroscopy analysis of the reactor wall surfaces shows that the inner parts of the Al2O3 chamber are strongly fluorinated (formation of Al–F bonds) during the SF6 plasma. At the same time the AlFx layer is sputtered from some parts of the chamber (mostly from the roof, which is bombarded by high energy ions), and AlF redeposition is observed on other parts of the reactor body. Hence, the cleaning process of the reactor leaves AlF residues on the chamber wall on its own. This leads to several issues including flake off of AlxFy particles on the wafer and process drifts (due both to the progressive growth of AlF material on the SiO2 windows and to the release of F atoms from the chamber walls during the etching process). This indicates that a strategy other than dry-cleaning the Al2O3 chamber walls in fluorine-based plasmas should be found. In this paper we have investigated two different strategies. The first one consists of replacing Al2O3 covering the chamber walls by another material for the chamber walls inner coating. In particular, we have investigated the surface modification of several types of organic polymers (Teflon, Parylene and carbon-rich polymers), when exposed to SF6-based plasmas. We show that these materials can be reset to their original condition after exposure to a dry-cleaning process because carbon containing polymers are slowly etched away by the SF6/O2 plasma. This suggests that the replacement of the conventional Al2O3 chamber wall material by a carbon-coated liner should be possible. Alternatively, we also propose a powerful strategy for conditioning and cleaning an Al2O3 reactor, in which a thin carbon-rich layer is deposited on the reactor walls by a short plasma step prior to any etching process. After etching, the SiOyClz layer deposited on the carbon layer during a silicon gate etch step can be cleared with an appropriate plasma, and the carbon layer removed by an O2 plasma, thus resetting the reactor walls to their initial state. Using this strategy the etching process always starts under the same chamber walls conditions (a carbon-rich wall) and is thus reproducible. At the same time, the issues associated with AlF deposits are prevented because the carbon-coated layer protects the Al2O3 chamber walls, and there is no fluorine released into the plasma. Finally, we will show that the etching profiles of the silicon gates and the selectivity towards the thin gate oxides are excellent in the carbon-coated chamber. This strategy is thus promising for future metal gate etching applications.
Read full abstract