A novel methodology of critical dimension statistical process control

SPC (Statistical Process Control) is a necessary tool for the current IC fabrication processing, especially when devices are shrinking down to 0.35 /spl mu/m and beyond. Most FABs have real time SPC systems to help process engineers to control process quality. In addition to real time SPC systems, some other process control methodologies and indices might be necessary for enhancement of process control monitoring. We have applied a statistical program using concepts of "hypotheses testing" and "max shift" to analyze variation and drift of CD (Critical Dimension). Based on collected CD data of the past 6 months, we conclude a methodology that could identify whether the CD variation is caused by drift of machine or process variation. To minimize process induced variation, we find lens heating compensation, dynamic focus and TARC application are effective. To respond to CD drift or shift caused by machine, we generate a daily report for the process engineer. Even without drawing the trend charts, process engineers can figure out the machine difference by the daily report. Combined with engineers' knowledge and know-how, the causes for CD variation can be identified much faster than before. As a result, the percentage of parameters with Cpk>1.33 increases by about 30%, compared with the Cpk performance before the statistical methodology was implemented.