Multiscale prediction of functional self-assembled materials using machine learning: high-performance surfactant...

Area of Science:

• Materials Science
• Computational Chemistry
• Physical Chemistry

Background:

• Controlling chemical molecular structures is key to tuning functional material properties.
• Machine learning (ML) combined with molecular simulation shows promise for predicting material electronic structures.
• Prior ML applications in materials science primarily used microscale data (molecular and electronic structures).

Purpose of the Study:

• To investigate the potential of ML for predicting multiscale data in functional materials.
• To explore ML's capability in predicting physical properties (dispersion, viscosity) from molecular structures of surfactants.
• To bridge the gap between molecular-level design and macroscopic material behavior.

Main Methods:

• Utilized machine learning models trained on chemical molecular structures of surfactants.
• Focused on predicting dispersion and viscosity as key physical properties of self-assembled surfactant solutions.
• Employed a self-assembly functional material system to test multiscale prediction capabilities.

Main Results:

• Machine learning accurately predicted physical properties such as dispersion and viscosity from surfactant molecular structures.
• Demonstrated that ML can effectively bridge different scales, from molecular structure to solution properties.
• Validated the feasibility of using ML for multiscale systems involving molecules, self-assembled structures, and bulk properties.

Conclusions:

• Machine learning offers a powerful tool for predicting multiscale systems in materials science.
• This study advances molecular design by enabling property prediction from fundamental chemical structures.
• The findings support broader applications of ML in developing novel functional materials.