Multiple layers’ effects on OCD measurement

OCD (Optical Critical Dimension) is a sufficiently accurate and reliable metrology used in semiconductor industry. Understanding how to setup a proper library will not only help to give the measurements that is needed, it will also prevent unnecessary downtime to the wafer fab factory. In this paper, a multiple layer stack, Poly-ONO-Poly, was seen performing according to the usual baseline until there was a downward shift seen on an OCD parameter SWA (Side Wall Angle). We took some readings off the suspected processes, ETCH trim time and tool's dependency and LITHO's resist thickness correlation to try to understand more on their effects on SWA. However, there is little correlation seen from them to the effects on SWA. Simulations were then done to show the spectra's relationship to the changes in the parameters: SWA, CD, HEIGHT. SWA was seen to have the least sensitivity to raw spectra versus CD and HEIGHT. This led to the suspicion that there was some changes in the spectra that caused the change in SWA, and thus leading us off the ETCH and LITHO into diffusion that could have changed the stack during ONO formation. It was thus found that ONO thickness has shifted. Once the thickness was tuned back to its original intent, SWA followed suit to recover to its measured units. This shows us that there's a need to fully understand the library/spectra involved when using OCD.