Online information source for semiconductor professionals

MARTINE RIBEYRON & MICHEL FUENTES, STMicroelectronics, Crolles, France ALAIN FERRERES, STMicroelectronics, Rousset, France

ABSTRACT

In the semiconductor field, the ITRS roadmap defines the guidelines for Facilities Fluids Quality, according to the technology node. However direct correlation between fluid purity and line widths/wafer yield is difficult because many other parameters must be considered such as: process recipes, process tools parameters etc. Although changes in technology can sometimes require higher fluid quality, the most daunting task for the Facilities group is to design fluids dispensing systems able to face rapid Fab start up with the required quality. To match with line width constraints, production Fabs are always benchmarking with R&D Fabs at the leading edge of line widths. We can submit some data and develop some lessons learned based on one new 8 inch R&D fab and one new 8 inch manufacturing fab. We can illustrate the criticality of the timing between line width, and fluid (gases, chemicals and ultra pure water) purity, to highlight that the leaching period is the key parameter for manufacturing fab, to analyse the risk during a busy period of yield improvement and new technology qualification. For an R&D fab the target of “no leaching period” is fast becoming a challenge to avoid fluids purity interference with development. We will give the example of the start-up and qualification of new chemical delivery systems (CDS) to illustrate this point. An overview of process fluid purity shows that average purity values must be distinguished from spot contamination events due to installation failures, or operational errors. The ability of facilities to avoid, or to anticipate purity excursions (robustness issue) is more important than the warranties offered by process system suppliers. Manufacturing and process personnel are comfortable with Facilities only if they can demonstrate a methodology to avoid excursions, whatever the commitments on specifications. We can also illustrate that even with strong methodology, extraordinary events do occur, which impact the yield. An operational field approach of the crisis seems more efficient than a theoretical one based only on specification.