This research aims to develop new methods for collecting and analyzing data related to defects and degradation of nano-sized devices. Specific problems in nano-manufacturing are addressed, such as reliability analysis of field emission displays and metal-oxide semiconductor (MOS) with sub 2 nm equivalent oxide thickness dielectrics. Statistical methods for modeling degradation and defects will include spatial point process modeling with change-points and random effects. Failure based models will consider nano-scale effects such as electron traps, interface states and the spatial relationship of defects across the oxide layer. The interdisciplinary research team includes expertise in chemical engineering, industrial engineering and statistics, and MOS devices will be fabricated as well as characterized for reliability issues. Experimental results will be used and verified in the development of theoretical models.
Results will create a more flexible framework for depicting the spatial distribution of defects (e.g., pair-potential Markov point process), which will be crucial for accurate reliability modeling. With this approach, state-of-the-art statistical models for nano-reliability will include hierarchical modeling, order-restricted inference, change-point regression, bias-adjusted bootstrap sampling and random effects modeling. The proposed study is physics-based from the PI's nano-electronics laboratories. Although nano-science has rapidly developed in chemistry and material science, reliability engineering and data analysis techniques for nano devices have not been sufficiently developed, especially in the United States. If successful, the outcome of this grant should help to change that, and the greatest impact could be the adaptation of advanced statistical analysis for nano-scale research. This research should also catalyze further statistical modeling in nanofabrication beyond the field of reliability.
