Machine learning inversion from scattering for mechanically driven polymers.

Area of Science:

• Polymer Physics
• Computational Materials Science
• Machine Learning Applications

Background:

• Understanding polymer behavior under mechanical stress is crucial for materials science.
• Scattering functions provide insights into polymer structure and dynamics.
• Extracting detailed parameters from scattering data can be computationally challenging.

Purpose of the Study:

• To develop a machine learning inversion method for analyzing scattering functions of mechanically driven polymers.
• To extract polymer feature parameters, including energy parameters and conformation variables.
• To validate the accuracy of the machine learning approach for polymer analysis.

Main Methods:

• Modeling polymers as chains with bending energy and external forces (stretching, shear).
• Generating a dataset of energy parameters, Monte Carlo scattering functions, and conformation variables.
• Utilizing Principal Component Analysis (PCA) to ensure machine learning feasibility.
• Training and validating a Gaussian Process Regressor (GPR) for parameter extraction.

Main Results:

• The machine learning inversion method successfully analyzes scattering functions.
• Feature parameters, including energy parameters (bending modulus, forces) and conformation variables (end-to-end distance, radius of gyration), were accurately extracted.
• The GPR model demonstrated effective validation on unseen data.

Conclusions:

• Machine learning, specifically GPR, provides a feasible and effective method for polymer parameter extraction from scattering data.
• This approach enhances the analysis of mechanically stressed polymer systems.
• The developed method offers a powerful tool for polymer characterization and simulation.