- Recipe
- Recipe setpoints and execution phases
- Offset
- Compensation or Offsets Used in This Run
- Notch
- The Wafer Orientation Notch
- DOE
- An Experimental Design That Deliberately Varies Conditions and Compares Results
Understanding Two Complementary Data Types Together
Edge defect concentration on an image is an outcome observation; recipes and sensors describe gas flows, RF power, and temperature conditions. Many sensors measure bulk chamber averages rather than localized pixels; optical color shifts cannot automatically be equated to precise film thickness.
- 01ImageFirst Document What Was Observed
Coordinates, morphology, inspection coverage, and metrology recipes.
- 02IdentityConfirm Associated Equipment from Actual History
Wafer ID, process step, chamber ID, and actual processing timestamps.
- 03ConditionsCross-Check Recipes and Process Logs
Stored setpoints, actual telemetry, controller compensations, and maintenance logs.
- 04RecommendationsFormulate Discriminative Verification Tests
Clarify which hypothesis tier is supported and which process conditions remain unverified.
Four Foundational Anchors Transforming Association into Comparability
- Entity:Confirm Wafer ID and physical tool/chamber, rather than joining solely by lot or cassette slot.
- Time:Distinguish track-in/out, chamber-processing, and inspection times, recording timezone, precision, and clock offsets.
- Space:Check origins, units, rotation, mirroring, notch orientation, and die/exposure-field definitions.
- Version:Track actual recipe version, run-time offsets, tool configuration, maintenance state, and inspection recipes.
A recipe hash identifies a saved file, but temporary offsets, controller overrides, or different parsing can change execution. A single hash does not fully describe actual conditions.
An Illustrative Investigation of Repeating Exposure-Field Anomalies
Assume anomalies recur across similar locations in multiple exposure fields. First verify repeating structures and inspection algorithms, then compare reticles, layout sensitivities, focus/dose logs, and underlying topography. Stepper repetition does not immediately eliminate photoresist or clean issues.
If focus records also differ, this supports further checking but does not establish a leaking chuck. Engineers can arrange applicable metrology, equipment checks, or controlled focus/dose trials to determine whether the difference explains the defects. This is not a real customer incident or a prescribed equipment repair procedure.
When Can Specific Adjustment Recommendations Be Proposed?
If an applicable response model, acceptable parameter range, and verification data exist for the current material and tool, an adjustment plan can be proposed for engineer review. Otherwise, recommend where to check, what data to add, and how to judge the result; do not recommend unvalidated production parameters directly from regression correlations or image morphology.
WAFERPILOT / APPLICATION TO THIS TOPIC
Placing Images, Process Conditions, and Recommendations into Traceable Reports
WaferPilot organizes connected images, existing analyses, and process context, selecting checks and recording comparative baselines according to method specs. Reports connect observations, algorithms/parameters, outputs, candidate mechanisms, and verification plans, preserving unverified data gaps.
Deliverables for Engineers:A report mapping visual phenomena to actual processing conditions, indicating supported hypothesis tiers and proposing physical verification or tuning plans.
View the Current Demo and Capability Scope →Related Reading & References
Below are verifiable industry references; specific processes, permissions, and acceptance criteria are governed by fab site specifications.