Supplementary Material

The supplementary material includes the full Python-based implementation of the AI-driven optimization framework described in the manuscript. It consists of three modules: 1. COMSOL Data Generator: A script that automates 500 high-fidelity COMSOL simulations using Latin Hypercube Sampling and extracts key plasma parameters such as electron density, uniformity, and absorbed power. 2. DNN Surrogate Model: A complete training pipeline using TensorFlow/Keras, including data preprocessing, model architecture, training, evaluation, and visualization of prediction accuracy. 3. Genetic Algorithm Optimization: A DEAP-based evolutionary optimization script that identifies optimal RF power and gas pressure values to maximize electron density while maintaining plasma uniformity above 90%. These scripts enable full reproducibility of the AI-enhanced plasma optimization workflow and can be adapted for other reactor geometries or plasma chemistries.