It has been found that semiconductor etch processes developed on eight and ten-inch barrels are not necessarily transferable to twelve-inch systems. This was found to be the case for Plasma Descum process. A study was undertaken to characterize such a process in a Branson/IPC L2101 Barrel System. Performance objectives included removing 500 – 1000 Angstroms while maintaining an Across the Wafer Uniformity (A.W. Unif.) and wafer to Wafer Uniformity (W.W. Unif.) of ± 10%.
This study was conducted with four-inch wafers exclusively. Three wafers (with dummy wafers in front and back) were used to locate optimum parameters. Once this was established, fifty wafer loads in different spacing arrangements were used to check for loading effects ana reproducibility. An Branson/IPC L2101 System with a 100 cfm Sargent Welch Pump was used on all tests. A Nitrogen plasma (2.0
Torr, 700W, 1 SLM) was used as a ~re-neat cycle on all runs except where noted otherwise. Oxygen was used as the process gas in the Descum cycle. Variable parameters included cycle time, pressure, RF power, Pre-heat temperature and gas flow. An anodized etch tunnel was used on all runs. See attached data sheet for specific listings.
It becomes quite apparent when looking at the Data Summary, that a new process was needed to improve upon the plasma descum process developed in smaller systems. Run #4 produced the desired results with an improvement of approximately 49% over the standard process (Run 11). Note, however, the loading effect and loss of uniformity when full loads were run (Runs 8-9). Thus, it was established that while postbaking an higher pre-heat cycles were found to be essential for improved uniformity, more process development work is needed for larger loads and larger substrate sizes.
However, these uniformity numbers are not accurately describe the condition of the processed samples. The uniformity rates quoted here are actually the rate removal uniformities. This describes how fast and now uniform the film is being “etcnea away”. The important spec. to consider is now uniform the remaining film is. This spec. gives a quantified description or the film before it goes into further process steps in the fabrication of that circuit. In discussing rate removal uniformities, a variation of 50 Angstroms in removing 500 Angstroms yields +20%! Yet, in looking at the film remaining, which can oe I.o to 1.5 microns in thickness, a variation of 200 Angstroms woula yield a uniformity of +2%! The uniformities of the film remaining in runs 8 and 9 were +1%. In most instances, this is at least equal to the uniformity of the film’s thickness before they are even descummed.
