Machine Vision for In-Situ Semiconductor Inspection
As semiconductor processes become more complex, manufacturers need to detect anomalies closer to where they occur. Machine vision can provide additional imaging information about the process, equipment, and substrate, complementing dedicated AOI and semiconductor metrology.
Why vision is moving further forward in the process
A semiconductor wafer or glass panel can pass through hundreds of process and handling steps. With advanced packaging, additional processes such as wafer thinning, fine-pitch RDL, bonding, debonding, and repeated substrate handling add further complexity.
As more value accumulates through these steps, late detection can mean higher substrate loss and greater financial impact. A damaged substrate entering process equipment may also cause substrate breakage, followed by tool cleaning, recovery, revalidation, and production downtime.
Traditional standalone inspection remains essential, particularly for high-resolution defect inspection and precision metrology. But it raises an additional question:
Can anomalies be detected earlier in the process step where they originate?
Machine vision can provide additional visual checkpoints during a process or immediately around it, helping engineers observe process events, equipment behavior, and substrate conditions earlier.
What does “in situ” mean from a machine vision perspective?
In situ is Latin for “in place.” In semiconductor manufacturing, it generally refers to measurements or observations made where the process occurs.
From a machine vision perspective, it is useful to distinguish two related tasks: process and tool monitoring, and additional wafer or panel inspection.
Process and tool monitoring
What is happening?
Machine vision captures dynamic events while the process or equipment is operating.
Examples include:
liquid dispensing
process behavior through a chamber window
robot movement
wafer or panel handling
The objective is to make process and equipment behavior visible for monitoring and troubleshooting.
Adding wafer or panel AOI
What happened to the substrate?
Vision can also inspect the wafer immediately before or after a process step, while it remains within the equipment flow.
Typical targets include:
coating anomalies
contamination
scratches
edge damage
This is equipment-integrated or process-near inspection rather than strict in-situ monitoring.
What can machine vision do close to the process?
The value of process-near vision is not always finding the smallest possible defect. Often, it is about capturing critical visual evidence at the right location and time.
The following two application cases show how machine vision can observe the process, equipment, and substrate at different points in semiconductor manufacturing.
Case 1: Monitoring liquid dispensing during CMP
Requirement
During chemical mechanical planarization (CMP), slurry is continuously supplied during polishing. Its flow behavior can affect material removal, defect formation, and polishing quality.
The requirement was to visually capture the liquid dispensing sequence while the process is taking place, rather than relying only on the wafer condition afterward.
Challenge
The dispensing event is dynamic and occurs while machine components are moving. A single image cannot show the complete sequence, while continuous recording creates unnecessary image data.
The imaging system therefore needs to capture the relevant section of the liquid-dropping process with sufficient temporal and image quality, synchronized to the machine sequence.
Solution
Two 2.3 MP GigE ace 2 area scan cameras are integrated to provide upper and lower views of the dispensing process. pylon vTools is used to trigger and control the acquisition sequence.
This allows the system to record the relevant process window for checking dispensing irregularities such as unexpected droplets, rather than continuously recording video.
See how the vision system was implemented
Continue reading for the imaging challenges, system setup, and solution approach, plus another application example of adding wafer AOI closer to the process.