Leveraging Physics Driven CMP Models in Manufacturing

摘要

Using extracted geometric design information, process models for CMP have been shown to accurately predict the systematic thickness variation across an as-manufactured chip [1,2,3]. This capability directly lends itself to the DFM related tasks of RC extraction and timing analyses, as well as sign off and hot spot detection for manufacturing [4,5,6,7,8,9]. In addition to these design side tools, a physics based modeling framework can provide valuable process development and optimization insight to the process engineer, before the process flow is fixed for a given technology node. By accurately extrapolating the effects of process changes, such as film stack thicknesses, polish times, and polish pressures, these CMP models can be used to efficiently optimize important process metrics, like throughput, planarity, and robustness to design and process variations, with fewer physical experiments. Purely empirical models cannot be used in this manner during process development, as they are implicitly tuned to a specific set of process conditions. This paper will demonstrate a number of useful CMP process optimization methodologies that are available when using physics driven CMP models with explicit process parameter settings.