Hybrid Metrology

Keywords: hybrid metrology, hm, metrology

Hybrid Metrology combines multiple measurement techniques to achieve accuracy beyond any single method — fusing data from different metrology tools (OCD, CD-SEM, AFM, TEM) using statistical methods to resolve each technique's blind spots, increasingly essential as single techniques hit physical limits at advanced semiconductor nodes.

What Is Hybrid Metrology?

• Definition: Integration of multiple metrology techniques for improved accuracy.
• Method: Collect measurements from different tools, fuse using statistical algorithms.
• Goal: Overcome limitations of individual techniques.
• Output: More accurate, comprehensive characterization than any single tool.

Why Hybrid Metrology Matters

• Single-Tool Limitations: Each technique has blind spots, biases, trade-offs.
• Accuracy Requirements: Advanced nodes demand sub-nanometer accuracy.
• Complex Structures: 3D structures (FinFET, GAA) challenge single techniques.
• Cross-Validation: Multiple techniques provide confidence in measurements.
• Cost-Effective Accuracy: Combine fast inline tools with accurate reference tools.

Metrology Technique Strengths & Weaknesses

OCD (Optical Critical Dimension):

• Strengths: Fast, non-destructive, multi-parameter, inline capable.
• Weaknesses: Model-dependent, limited resolution, averaging over measurement spot.
• Best For: High-throughput monitoring, trend tracking.

CD-SEM (Critical Dimension SEM):

• Strengths: High resolution, direct imaging, edge detection.
• Weaknesses: Top-down view only, charging effects, slow.
• Best For: CD measurement, pattern inspection.

AFM (Atomic Force Microscopy):

• Strengths: True 3D profile, sidewall measurement, no charging.
• Weaknesses: Very slow, tip convolution, limited throughput.
• Best For: Reference metrology, sidewall angle, 3D structures.

TEM (Transmission Electron Microscopy):

• Strengths: Highest resolution, cross-section view, material contrast.
• Weaknesses: Destructive, extremely slow, expensive, sample prep.
• Best For: Gold standard reference, failure analysis.

Hybrid Metrology Approaches

OCD + CD-SEM:

• Combination: OCD for multi-parameter + SEM for absolute CD calibration.
• Method: Use SEM to calibrate OCD model, then use OCD for production.
• Benefit: OCD speed with SEM accuracy.
• Application: Lithography and etch process control.

OCD + AFM:

• Combination: OCD for throughput + AFM for 3D profile validation.
• Method: AFM validates sidewall angle, OCD uses for production.
• Benefit: 3D accuracy with optical speed.
• Application: Complex 3D structures, FinFET, GAA.

CD-SEM + AFM:

• Combination: SEM for top CD + AFM for height and sidewall.
• Method: Fuse top-down and 3D information.
• Benefit: Complete 3D characterization.
• Application: Resist profile, etch profile characterization.

Multi-Tool + TEM Reference:

• Combination: All inline tools calibrated against TEM.
• Method: TEM provides ground truth for model validation.
• Benefit: Traceable accuracy to highest standard.
• Application: New process development, metrology qualification.

Data Fusion Methods

Weighted Average:

• Method: Combine measurements weighted by uncertainty.
• Formula: x_fused = Σ(w_i · x_i) / Σ(w_i), where w_i = 1/σ_i².
• Simple: Easy to implement and understand.
• Limitation: Assumes independent, unbiased measurements.

Bayesian Fusion:

• Method: Combine measurements using Bayesian inference.
• Prior: Incorporate prior knowledge about parameters.
• Posterior: Update beliefs based on all measurements.
• Benefit: Principled uncertainty quantification.

Machine Learning Fusion:

• Method: Train ML model to predict true value from multiple measurements.
• Training: Use reference metrology (TEM) as ground truth.
• Benefit: Learns complex relationships, handles biases.
• Challenge: Requires substantial training data.

Kalman Filtering:

• Method: Sequential fusion with temporal correlation.
• Application: Combine measurements over time.
• Benefit: Optimal for time-series data.

Benefits of Hybrid Metrology

Improved Accuracy:

• Uncertainty Reduction: Fusing N measurements reduces uncertainty by ~√N.
• Bias Cancellation: Different techniques have different biases.
• Cross-Validation: Inconsistencies reveal measurement issues.

Comprehensive Characterization:

• Multiple Parameters: Each technique measures different aspects.
• 3D Information: Combine top-down and cross-section views.
• Material Properties: Optical + physical measurements.

Cost-Effective:

• Sparse Reference: Expensive techniques used sparingly for calibration.
• Inline Speed: Fast techniques for production monitoring.
• Optimal Resource Use: Right tool for right purpose.

Robustness:

• Redundancy: If one technique fails, others provide backup.
• Outlier Detection: Inconsistent measurements flagged.
• Confidence: Multiple techniques increase confidence.

Implementation Framework

Reference Metrology:

• Gold Standard: Establish TEM or AFM as reference.
• Calibration: Calibrate inline tools against reference.
• Frequency: Periodic recalibration (weekly, monthly).

Inline Monitoring:

• Primary Tool: Fast technique (OCD, SEM) for production.
• Sampling: High-frequency measurements.
• Feedback: Real-time process control.

Statistical Fusion:

• Algorithm: Implement fusion algorithm (weighted average, Bayesian, ML).
• Uncertainty: Propagate uncertainties through fusion.
• Output: Fused measurement with confidence interval.

Validation:

• Cross-Check: Compare fused results with reference.
• Residual Analysis: Check for systematic errors.
• Continuous Improvement: Refine fusion algorithm over time.

Challenges

Tool-to-Tool Matching:

• Systematic Offsets: Different techniques may have biases.
• Calibration: Requires careful cross-calibration.
• Drift: Tools drift over time, need periodic recalibration.

Data Integration:

• Different Formats: Each tool has different output format.
• Spatial Registration: Measurements at same location.
• Timing: Synchronize measurements in time.

Computational Complexity:

• Real-Time: Fusion must be fast enough for inline use.
• Algorithm: Balance accuracy vs. computational cost.
• Infrastructure: Requires data management system.

Cost:

• Multiple Tools: Requires investment in multiple metrology platforms.
• Maintenance: More tools to maintain and calibrate.
• Training: Staff must understand multiple techniques.

Applications at Advanced Nodes

FinFET Metrology:

• Challenge: 3D structure with critical dimensions in all directions.
• Solution: OCD for fin pitch + AFM for fin height + SEM for fin width.
• Benefit: Complete 3D characterization.

GAA (Gate-All-Around):

• Challenge: Nanowire/nanosheet dimensions, buried structures.
• Solution: Hybrid OCD + X-ray + TEM for validation.
• Benefit: Non-destructive monitoring with TEM validation.

EUV Patterning:

• Challenge: Stochastic effects, LER/LWR, defects.
• Solution: SEM for LER + OCD for CD + AFM for 3D profile.
• Benefit: Comprehensive patterning quality assessment.

Tools & Platforms

• KLA-Tencor: Integrated hybrid metrology solutions.
• ASML: YieldStar + e-beam hybrid metrology.
• Nova: Integrated OCD + SEM systems.
• Bruker: AFM for hybrid metrology reference.

Hybrid Metrology is essential for advanced semiconductor manufacturing — as single metrology techniques reach their physical limits, combining multiple methods through intelligent data fusion provides the accuracy, comprehensiveness, and confidence required for process control at 7nm and below, making it indispensable for next-generation semiconductor fabrication.

Source: ChipFoundryServices — Search this topic — Ask CFSGPT