Method for process monitoring with optical inspections
Machine learning approaches provide additional information about semiconductor wafer inspection stability issues that makes it possible to distinguish consequential process variations like process excursions from minor process variations that are within specification. The effect of variable defect of interest (DOI) capture rates in the inspection result and the effect of variable defect count on the wafer can be monitored independently.
1. A method comprising:
optically inspecting a plurality of wafers with an optical inspection tool thereby generating images;
extracting features from the images with machine learning-based classifiers using a processor;
determining, using the processor, monitoring metrics from the features and a population of all detected events for the wafer;
receiving, at the processor, classified defects of the wafers from a scanning electron microscope;
determining, using the processor, separability metrics for the classified defects in relation to inspection thresholds, wherein the separability metrics account for a population distribution of the classified defects; and
determining, using the processor, separability trends of the defects for the wafers.
2. The method of claim 1 , wherein the optical inspection tool is a broad band plasma inspection tool.
3. The method of claim 1 , further comprising filtering nuisance from the detected events for the wafer using the processor.
4. The method of claim 3 , further comprising determining confidence values for each of the detected events for the wafer using the processor.
5. The method of claim 4 , further comprising determining, using the processor, defect movement with respect to inspection thresholds.
6. The method of claim 5 , wherein determining the defect movement includes:
projecting all the detected events on a confidence axis; and
monitoring motion of the classified defects along the confidence axis with respect to a threshold.
7. The method of claim 5 , wherein determining the defect movement includes classifying the defect movement from process changes that impact yield and process changes that affect inspection sensitivity.
8. The method of claim 1 , wherein the separability trends monitor confidence that classified defects are captured.
9. The method of claim 8 , wherein the separability trends monitor signal strength.
10. The method of claim 8 , wherein the separability trends monitor signal spread.
11. A non-transitory computer readable medium storing a program configured to instruct a processor to execute the method of claim 1 .
12. A system comprising:
an optical inspection tool including:
a light source
a stage configured to hold a wafer; and
a detector; and
a processor in electronic communication with the detector, wherein the processor is configured to:
generate images of a plurality of the wafers using data from the detector;
extract features from the images with machine learning-based classifiers using a processor;
determine monitoring metrics from the features and a population of all detected events for the wafer;
receive classified defects of the wafers from a scanning electron microscope;
determine separability metrics for the classified defects in relation to inspection thresholds, wherein the separability metrics account for a population distribution of the classified defects; and
determine separability trends of the defects for the wafers.
13. The system of claim 12 , wherein the light source is a broad band plasma source.
14. The system of claim 12 , wherein the processor is further configured to filter nuisance from the detected events for the wafer.
15. The system of claim 14 , wherein the processor is further configured to determine confidence values for each of the detected events for the wafer.
16. The system of claim 15 , wherein the processor is further configured to determine defect movement with respect to inspection thresholds.
17. The system of claim 16 , wherein determining the defect movement includes:
projecting all the detected events on a confidence axis; and
monitoring motion of the classified defects along the confidence axis with respect to a threshold.
18. The system of claim 16 , wherein determining the defect movement includes classifying the defect movement from process changes that impact yield and process changes that affect inspection sensitivity.