YIELD, STATISTICS AND JMP Yield JMP: a Statistical Tool for Yield Enhancement in Semiconductor Industry
Francois Bereget, Ippon Innovation; Alexandre Courvat, Soitec
Semiconductor is a very complex industry, dealing with hundred of operations (process stages) and hundred of parameters, with products sensitive to particles as small as 0.1 microns, and leading edge technologies at work. SOITEC is manufacturing substrate wafers based on a technology called SOI - Silicon On Insulator, providing key chipmakers with raw material allowing low consumption and high performance. In a large volume wafer fab, with very strong customer requirements in terms of quality, new process control tools are necessary in order to quickly detect any faulty process tool, and also to find root cause of complex yield issues related to some process stages, but not tool dependant. The key requirement from the yield engineers is a statistical tool with two characteristics: automatic enough to quickly highlight a very limited set of process parameters (among hundred) responsible for yield losses, interactive enough to be able to tune the analysis taking into account “engineering knowledge“. JMP software was selected for 3 reasons: it is using advanced statistical method adapted to semiconductor complexity, it is very user friendly and all engineers can perform additional stats with no pain, it includes a powerful programming language, JSL, to develop automatic analysis for thousand of input and output parameters.