R2R Automatic Tuning: How Run-to-Run Control Improves Wafer Yield
Key Takeaway
R2R automatic tuning improves wafer yield by 3-8% through closed-loop process control. The system uses VM predictions to calculate compensation values and automatically adjusts recipe parameters for each wafer run, eliminating manual tuning delays.
The Era of Manual Tuning Is Coming to an End
In semiconductor manufacturing, precise control of process parameters directly determines wafer yield. The traditional approach relies on experienced process engineers who manually adjust equipment process parameters (recipes) based on metrology results and personal judgment to compensate for equipment drift and process variation.
This approach has significant limitations:
- Slow response: From the time an anomaly is detected to the completion of parameter adjustment, dozens or even hundreds of wafers may have already been processed
- People-dependent: The experience of senior engineers is difficult to standardize and transfer; personnel turnover directly impacts yield
- Imprecise: Manual tuning typically amounts to “coarse adjustment,” making it difficult to achieve fine-grained wafer-by-wafer optimization
R2R Automatic Tuning: From Human Experience to Algorithmic Closed-Loop Control
R2R (Run-to-Run) control is an advanced process control strategy — after each wafer processing run is completed, feedback data from that run is used to automatically calculate and adjust the process parameters for the next run. It elevates process optimization from “human experience-driven” to a “data and algorithm-driven” automatic closed loop.
The core R2R cycle operates as follows:
- Processing: The equipment executes the process with current parameters
- Feedback: Quality results are obtained through Virtual Metrology (VM) or physical metrology
- Computation: The R2R algorithm calculates parameter compensation based on the deviation between target and actual values
- Adjustment: New parameters are automatically dispatched to the equipment for the next processing run
R2R Applications in Critical Process Steps
Chemical Mechanical Polishing (CMP)
In CMP processes, polishing pad wear and slurry concentration changes cause removal rate drift. R2R compensates in real time by adjusting polishing time or pressure parameters, ensuring consistent film thickness across every wafer.
Etch
During plasma etching, chamber conditions change over time. R2R can automatically adjust RF power, gas flow rates, and other parameters to compensate for the impact of equipment state changes on etch depth and uniformity.
Thin Film Deposition
In CVD/PVD processes, target consumption and temperature drift affect film thickness. R2R uses closed-loop parameter adjustment to keep film thickness variation within an extremely narrow process window.
VM + R2R: The Most Powerful Combination
The effectiveness of R2R is highly dependent on the timeliness and coverage of feedback data. If R2R relies solely on physical metrology (with a sampling rate of 5%-10%), both the tuning frequency and precision are constrained.
When VM is combined with R2R, the processing outcome of every single wafer can be predicted in real time, enabling the R2R algorithm to achieve wafer-to-wafer or even lot-to-lot fine-grained control, significantly improving yield stability.
SECS/GEM: The Communication Foundation for R2R Deployment
R2R automatic tuning requires bidirectional communication with equipment: reading equipment status and dispatching parameter changes. This relies on the semiconductor industry standard SECS/GEM protocol. Without a standardized equipment communication interface, R2R cannot achieve true “automatic” closed-loop control.
How to Get Started with R2R Deployment
R2R does not require a full-scale, immediate overhaul of the entire production line. The recommended approach is:
- Select one critical process step (e.g., CMP) to first validate R2R effectiveness
- Deploy a VM model to ensure real-time quality feedback
- Start with a conservative strategy, gradually expanding the R2R tuning range
MST Semiconductor’s NeuroBox Edge Intelligence Platform supports integrated VM + R2R deployment, enabling a complete closed loop from data acquisition and quality prediction to parameter optimization at the equipment level.
Related Reading
Deploy real-time AI process control with sub-50ms latency.
