Design of Experiments (Doe) for Semiconductor

Design of Experiments (DOE) in semiconductor manufacturing is a systematic, statistical methodology for varying process parameters to determine their effects on output quality — identifying which factors matter most and finding optimal operating conditions with the minimum number of experimental runs.

Why DOE Instead of One-Factor-at-a-Time (OFAT)?

• OFAT changes one variable while holding others constant. It requires many runs, misses interaction effects, and may find a local optimum rather than the true optimum.
• DOE changes multiple variables simultaneously in a structured pattern. It requires fewer runs, reveals interactions, and maps the full response landscape.
• A DOE with 5 factors and 2 levels per factor needs only 16–32 runs. OFAT testing the same factors might need 100+ runs to get equivalent information.

DOE Process in Semiconductor Context

• Define Factors: Select the process parameters to study (e.g., RF power, pressure, gas flow, temperature, time).
• Define Levels: Choose the range for each factor (e.g., power: 200W and 400W; pressure: 20 mTorr and 50 mTorr).
• Define Responses: What output to measure (e.g., etch rate, CD, uniformity, selectivity).
• Choose Design: Select appropriate DOE type (full factorial, fractional factorial, RSM, etc.).
• Run Experiments: Process wafers according to the DOE matrix — each run uses a specific combination of factor levels.
• Analyze Results: Use ANOVA, regression, and response surface analysis to determine which factors and interactions are statistically significant.
• Optimize: Find the factor settings that optimize the response(s).

Common Semiconductor DOE Applications

• Etch Recipe Development: Optimize etch rate, selectivity, profile, and uniformity simultaneously by varying power, pressure, gas flows, and temperature.
• Lithography Optimization: Find optimal dose, focus, PEB temperature, and develop time for best CD and process window.
• Deposition Tuning: Optimize film thickness, uniformity, stress, and composition.
• CMP Optimization: Balance removal rate, uniformity, dishing, and defectivity.
• Reliability Testing: Identify factors affecting device lifetime and failure modes.

Key DOE Concepts

• Main Effect: The direct impact of changing one factor on the response.
• Interaction Effect: When the effect of one factor depends on the level of another factor.
• Replication: Running the same condition multiple times to estimate experimental error.
• Randomization: Running experiments in random order to prevent systematic biases.

DOE is the essential methodology for semiconductor process development — it converts expensive, time-consuming trial-and-error into efficient, statistically rigorous optimization.