Robust Ion-Implantation Process Design Through Statistical Analysis

摘要

This paper demonstrates the importance of taking statistical process variations into account when designing semiconductor processes, and provides a methodology for doing so. An on-axis, high-energy ion-implant process is used as a vehicle to demonstrate the methodology. UT-MARLOWE is the well-calibrated (and hence predictive) simulation tool of choice. It is shown that crystal-cut errors and implanter-tilt errors can result in a considerably off-axis implant for this nominally on-axis implant: instead of the intended tilt = 0°, the implant actually occurs at an average tilt of ～0.65°, with a significant probability of having tilt > 1°. Knowledge of the error-statistics is used to design a robust process (i.e. a process relatively immune to statistical process variations), thereby providing better control over device performance. This methodology can also account for deterministic errors arising from beam divergence and process-disk geometry in multi-wafer machines.