Semiconductor manufacturing is a complex process with hundred of subprocesses. Different sensors record diverse information including the profiles of tools used in each fabrication subprocess, temperatures, pressures, etc. The output of this process is a semiconductor wafer. Any malfunctioning subprocess can introduce a yield loss in the output wafer and hence render the entire batch useless resulting in potentially millions of dollars of loss. It is therefore imperative to identify the subprocess responsible for the yield loss.
Our scientists have developed a tool to identify the malfunctioning subprocess out of hundreds of subprocesses in the fabrication process. We use the recorded data and have designed some customized algorithms using their geometrical insight on the collected data, and employing our strong back ground in scientific computing to pin-point the malfunctioning subprocess and reduce the scope of troubleshooting and human errors. This results in considerable reduction in the production of low-yield batches and can save potentially millions of dollars for our clients in the semiconductor industry.