Extreme Gradient Boosting to Predict Atomic Layer Deposition for Platinum Nano-Film Coating.

Area of Science:

• Materials Science and Engineering
• Artificial Intelligence and Machine Learning
• Chemical Engineering

Background:

• Atomic Layer Deposition (ALD) is crucial for precise thin-film fabrication.
• Controlling ALD process parameters is key to achieving desired material properties.
• Artificial intelligence offers advanced methods for process optimization.

Purpose of the Study:

• To apply the Extreme Gradient Boosting (XGBoost) algorithm to optimize platinum (Pt) nano-film production via ALD.
• To develop a predictive model for the Al/Pt component ratio based on ALD process parameters.
• To evaluate the performance of XGBoost against other machine learning models.

Main Methods:

• Platinum (Pt) was deposited on α-Al2O3 substrates using rotary-type ALD equipment.
• Four key process parameters (temperature, stop valve time, precursor pulse time, reactant pulse time) were varied.
• Inductively Coupled Plasma Atomic Emission Spectrometry (ICP-AES) was used to determine the Al/Pt component ratio.
• A dataset of 625 samples was generated, split into 500 training and 125 testing samples for XGBoost model development.

Main Results:

• XGBoost achieved an accuracy of 99.9% and a coefficient of determination (R²) of 0.99 in predicting the Al/Pt component ratio.
• The XGBoost model demonstrated a lower inference time compared to Random Forest Regression.
• Prediction safety was found to be higher with XGBoost than with Light Gradient Boosting Machine.

Conclusions:

• XGBoost is a highly effective AI algorithm for optimizing ALD processes, specifically for platinum nano-film deposition.
• The developed XGBoost model provides accurate and efficient predictions of the Al/Pt component ratio.
• XGBoost offers superior performance in terms of accuracy, speed, and prediction reliability for ALD applications.