Automatic problem detection and documentation in a plasma etch reactor

IC manufacturing involves a large number of complex process steps. There is a probability of a microprocessing error in each of these steps. The author describes two software tools, in a single-wafer plasma etcher, that have been developed to minimize the occurrence and maximize the efficiency of the diagnosis of such microprocessing. The first examines the endpoint trace of every wafer and determines whether that wafer has seen anomalous processing. If so, the software can terminate the processing of subsequent wafers. The second records the analog values of all the process control parameters during the entire etch process. When these two routines are linked appropriately, it is possible to record these analog values for only those wafers that have seen anomalous processing. This feature provides data for the analysis of the process problem. These procedures thus act to minimize the number of wafers that are microprocessed and provide pertinent diagnostic information for those that have been