Machine Learning Framework for Characterizing Processing-Structure Relationship in Block Copolymer Thin Films.

Area of Science:

• Materials Science
• Polymer Science
• Data Science

Background:

• Block copolymer (BCP) morphology dictates material properties and applications.
• Characterizing BCP thin films is crucial for material development.
• Existing methods can be time-consuming and lack high-throughput capabilities.

Purpose of the Study:

• To develop a machine learning (ML)-enabled, high-throughput framework for BCP thin film morphology analysis.
• To integrate grazing incidence small-angle X-ray scattering (GISAXS) and atomic force microscopy (AFM) data.
• To enable efficient exploration and optimization of BCP processing parameters.

Main Methods:

• A convolutional neural network (CNN) was trained to classify AFM images with 97% accuracy.
• High-throughput extraction of 2D grain size measurements from classified AFM images.
• ML models were trained to predict domain orientation based on processing parameters (solvent ratio, additive type/ratio).
• SHapley Additive exPlanations (SHAP) were used for model interpretability.

Main Results:

• GISAXS-based property predictions showed strong performance (R² > 0.75).
• AFM-based property predictions were less accurate (R² < 0.60) due to localized measurements.
• SHAP analysis identified additive ratio as the most impactful parameter for morphological predictions.
• The framework successfully correlated processing parameters with BCP morphology.

Conclusions:

• The developed ML framework accelerates the characterization of BCP thin film morphology.
• Interpretability analysis provides crucial insights into structure-property relationships.
• This approach facilitates the optimization of BCP materials by understanding parameter importance.
• The study offers a pathway for exploring and optimizing BCP morphology across a wide processing landscape.