As you correctly note, you probably could use an etch procedure to remove the Cu. For whatever reason chemically (see below for some speculations), the etch procedure is not very efficient compared to the CMP procedure. Presently, most effort is going into using CMP to remove the Cu.
(Chemical speculations) I would guess the poor etching of Cu is due to lousy kinetics and/or surface chemistry. The plasma environment is there to essentially generate high energy species that can then react (ex. Cu + Cl* - CuCl, Cu* + Cl2 - CuCl + Cl*, etc. (* is an activated species)). First, the reaction may be too slow to be economical. Secondly, it is possible that this type of etching is very nonuniform. For example, when the etching procedure begins and one moves through the top layer of Cu, some areas will be etched slightly faster than other areas. These surfaces where etching started slightly faster will have a different shape and may enhance the removal of the next layer of Cu. As a result, different portions of the Cu will be removed at vastly different rates and a very uneven etch pattern would occur. This would clearly be unacceptable for semiconductors. We use mica for various experiments we do, and have done some etching studies of mica using both hydroxyl radical (essentially a water-based plasma) and acid based etching. When one examines the resulting surface, some areas of the mica are very deeply etched (>30 molecular layers) while other areas are untouched and remain atomically flat. (Again, I want to stress that these are just speculations since I am not familiar with the chemistry of Cu.) |
