Virtual metrology is the practice of estimating what a metrology tool would measure by using process, sensor, and equipment data instead of physically measuring every part or wafer. It is especially common in semiconductor manufacturing, where direct measurement can be slow, expensive, and capacity-constrained.
Why It Exists
Physical metrology is critical, but it is not always practical to run it on everything. A fab may need a faster signal about thickness, critical dimension, overlay, or other process outputs than the metrology tools alone can provide. Virtual metrology helps fill that gap by offering a model-based estimate between or alongside direct measurements.
Why It Matters In AI
AI makes virtual metrology more useful because it can learn relationships between tool behavior, recipe settings, sensor traces, and final measurements. That lets manufacturers reduce some measurement overhead while still keeping a tighter feedback loop. In practice, virtual metrology often feeds advanced process control by giving the controller a faster estimate of current process quality.
What To Keep In Mind
Virtual metrology is not a replacement for physical metrology. It is a complement. The model still needs calibration, drift checks, and enough real measurements to stay honest. It works best when teams understand where the estimate is reliable, where uncertainty grows, and when direct measurement is still necessary.
Related Yenra articles: Micro-Fabrication Process Control and Semiconductor Defect Detection.
Related concepts: Advanced Process Control (APC), Predictive Analytics, Surrogate Model, Uncertainty, and Model Monitoring.