Key points are not available for this paper at this time.
A digital twin is an integrated multi-physics, multiscale, probabilistic simulation of a system that uses the best available physical models, sensor updates, process history, etc., to mimic and to predict the performance of the system. In the semiconductor industry, there is a need to establish such fundamental, mechanism-based, correlations between process conditions and observed process performance. Such understanding can lead to the creation of industry-wide technical standards, and digital twins (i.e., real time digital replicas of unit processes), that will enable inter-operability, ease process selection and integration, as well as help reduced cost of ownership. In this paper, we revisit our decades old, but still relevant, quest for appropriate mechanism-based metrology to support the digitization of Chemical-Mechanical Polishing (CMP) for the back-end-of line (BEOL), with advanced packaging, specifically hybrid bonding for 3D interconnects, in mind. Here, we consider the digitization of the nanoscale tribological aspects of CMP. For example, many CMP processes use chemically unstable polyurethane pads in a variety of slurry formulations which make the process difficult to understand and to model. We had previously described how the chemical instability of the PU-pads affects the CMP process performance. Here we contextualize the historical performance data, identify data gaps, and propose some next steps towards digital twin CMP modules.
Yaw S. Obeng (Fri,) studied this question.