Fragility in glassy liquids: A structural approach based on machine learning

Area of Science:

• Condensed matter physics
• Materials science
• Computational chemistry

Background:

• Glassy liquids exhibit a universal rapid increase in viscosity upon cooling.
• Liquid fragility, describing the deviation from Arrhenius behavior, varies nonuniversally and is poorly understood.
• Understanding the origins of liquid fragility is crucial for predicting material properties.

Purpose of the Study:

• To investigate the factors controlling the wide range of fragility observed in glassy liquids.
• To identify a universal structural descriptor correlated with dynamical rearrangements in supercooled liquids.
• To elucidate the relationship between structure and dynamics in glass-forming systems.

Main Methods:

• Simulated a family of harmonic sphere models spanning a wide range of fragility.
• Employed machine learning (support vector machine) to identify a structural order parameter termed "softness".
• Analyzed the correlation between softness and dynamical properties like relaxation time and viscosity.

Main Results:

• Identified "softness," a linear combination of structural quantities, as a universal order parameter across the studied liquids.
• Demonstrated that softness is highly correlated with dynamical rearrangements and relaxation times.
• Showcased that softness effectively distinguishes between strong and fragile liquid behaviors.

Conclusions:

• Softness is a key structural indicator that explains the varying fragility of glass-forming liquids.
• Machine learning provides a powerful tool for uncovering fundamental structure-dynamics relationships.
• This work advances the understanding of the microscopic origins of glass transition phenomena.