Optical emission spectral analysis using FastICA for diagnosing wafer edge uniformity in plasma etching

Area of Science:

• Plasma Physics and Chemistry
• Spectroscopy
• Data Analysis

Background:

• Spatially integrated optical emission spectroscopy (OES) signals obscure localized reaction dynamics, particularly at wafer edges.
• Existing methods for diagnosing wafer edge uniformity often require specialized hardware, increasing complexity and cost.

Purpose of the Study:

• To develop a computational method for resolving localized reaction dynamics at wafer edges using standard OES.
• To enhance wafer edge uniformity diagnostics without requiring hardware modifications.

Main Methods:

• Application of Independent Component Analysis (ICA), specifically the FastICA algorithm, to standard plasma emission spectra.
• Utilizing second-order temporal derivative analytics to isolate transient dynamics and oscillatory signatures.
• Correlation analysis between argon emission line intensity modulation and shifts in the electron energy distribution function (EEDF).

Main Results:

• ICA successfully decoupled edge-specific transient dynamics from the bulk plasma background.
• A distinct oscillatory signature linked to edge-localized impedance mismatch was identified.
• Intensity modulation of argon lines was found to correlate with EEDF shifts.
• A predictive model achieved a cross-validated coefficient of determination (Q²) of 0.913, demonstrating high accuracy.

Conclusions:

• The proposed ICA-based method offers a reliable and robust computational alternative to dedicated spatial sensors for wafer edge uniformity diagnostics.
• This approach effectively transforms standard 1D optical sensors into spatially resolvable diagnostic tools.
• The findings enable non-invasive, in-situ monitoring of critical wafer edge processes.