Molecular cluster analysis using local order parameters selected by machine learning

Area of Science:

• Materials Science
• Computational Chemistry
• Soft Matter Physics

Background:

• Understanding molecular structure is key to phase transitions.
• Local Order Parameters (LOPs) characterize local molecular arrangements.
• Automated methods are needed to screen numerous LOPs.

Purpose of the Study:

• Develop a machine learning package (MALIO) to automatically identify optimal LOPs.
• Apply MALIO to analyze liquid crystal phase transitions (nematic-smectic).
• Evaluate the performance of different LOPs and neighbor selection methods.

Main Methods:

• Developed the Molecular Assembly structure Learning package for Identifying Order parameters (MALIO).
• Utilized machine learning for automated LOP screening.
• Applied MALIO to uniaxial liquid crystal systems to study phase transitions.

Main Results:

• MALIO efficiently proposed optimal LOPs for distinguishing nematic and smectic phases.
• Identified LOP candidates for precise observation of nematic-smectic transitions.
• Examined the influence of LOP species and neighbor selection on local structure statistics.

Conclusions:

• Machine learning, via MALIO, effectively screens a vast number of LOPs.
• MALIO successfully identifies promising LOP candidates for materials analysis.
• The study demonstrates MALIO's capability in selecting the best LOP species for specific applications.