Abstract :
The problem discussed is as follows. As the oxide thickness is reduced, due to both the improvement of our knowledge on silicon dioxide breakdown and to the fact that industrials want to push the silicon dioxide to its limit, new approaches have been discussed during the last few years: The first one consists of the change in the extrapolation model: the "E" model is demonstrated to be pessimistic and the data better fitted by a power law model(~Vn). The second one deals with the link between oxide failure and circuit failure. It is undoubtedly demonstrated in the literature that at least some circuits can accept some breakdown spots without any significant modification of the circuit characteristics. The general questions are: are both approaches practically used? Based on which validations? What are the physical justifications? And finally, what is the oxide thickness limit when considering reliability specification? It was concluded that the introduction of new approaches is more and more required to fulfill the reliability specification with respect to breakdown for ultra-thin oxides. However, the complete understanding of the validity range of the power law model is not clear due to lack of a physical model and data collection on different technologies and structures. However, most of the work concerned the definition of new criteria and the prediction model to take into account the fact that the oxide breakdown occurrence can often be not be considered as a circuit failure. Thus, it seems that we are far from knowing if the limit in the decrease of nitrided silicon dioxide thickness is due to reliability or to its leakage.
