While in-line semiconductor metrology is very good at imaging and identifying problems in processes, it’s actually very bad at helping to address those problems before they have significant financial impact.
Beyond identification, what semiconductor processes need more of is in-situ, real-time quantitative analysis, to mitigate issues before they turn up as defects or excursions in electrical performance or line yield.
In other words, semiconductor process control needs to move from in-line monitoring to in-situ process measurement and control. Metrology needs to move from ‘we’ve found a processing problem, what do we do about it?’ to ‘we know our process is running right now within optimal limits and at the highest throughput efficiency’.
But how can this change be achieved in practice?