Exploring the landscape of Buckingham potentials for silica by machine learning: Soft vs hard interatomic forcefields

Area of Science:

• Materials Science
• Computational Chemistry
• Condensed Matter Physics

Background:

• Interatomic force fields are crucial for simulating silicate glasses, using partial charges to model ionic interactions.
• Force fields are categorized as
• soft
• or
• hard
• based on the partial charge assigned to silicon (Si) atoms.

Purpose of the Study:

• To employ machine learning to systematically explore the accuracy landscape of Buckingham force fields for silica.
• To identify and analyze the distinct characteristics of "soft" and "hard" force fields in predicting silica structures.

Main Methods:

• Utilized machine learning to efficiently map the accuracy of Buckingham force fields across a range of parameters for silica.
• Analyzed the structural properties, specifically short-range and medium-range order, of silica configurations generated by different force fields.

Main Results:

• Discovered that "soft" and "hard" force fields represent two separate, competitive local minima in the force field accuracy landscape.
• Observed that both force field types exhibit comparable accuracy for short-range order in silica.
• Determined that "soft" force fields demonstrate superior performance in describing the medium-range order of silica structures.

Conclusions:

• The study identifies distinct "soft" and "hard" Buckingham force field regimes for silica, each with unique structural prediction capabilities.
• "Soft" force fields offer an advantage in accurately representing medium-range order, which is critical for understanding the complex structure of silicate glasses.