Radial distribution function for liquid gallium from experimental structure factor: a Hopfield neural network...

Area of Science:

• Condensed Matter Physics
• Computational Materials Science
• Statistical Mechanics

Background:

• Determining the radial distribution function (RDF) is crucial for understanding liquid structures.
• Experimental structure factors provide essential data for RDF calculations.
• Traditional methods like Fourier transforms and Monte Carlo simulations have limitations.

Purpose of the Study:

• To investigate the efficacy of a Hopfield neural network (HNN) for retrieving the RDF of liquid gallium.
• To compare HNN performance against established methods using experimental data.
• To evaluate HNN's robustness under different initial conditions.

Main Methods:

• Utilized a Hopfield neural network for the inverse problem of RDF calculation.
• Employed experimental structure factor data for liquid gallium at 959 K.
• Tested HNN with two distinct initial conditions: ideal gas RDF and a square-well potential gas simulation.

Main Results:

• The Hopfield neural network accurately reproduced the radial distribution function of liquid gallium.
• HNN results were comparable to those obtained via Fourier transform and Monte Carlo simulations.
• Both tested initial conditions yielded accurate inverse results, demonstrating HNN's versatility.

Conclusions:

• The Hopfield neural network is a powerful and accurate tool for calculating RDF from experimental structure factor data.
• HNN offers a viable alternative to conventional methods for liquid structure analysis.
• This approach enhances the ability to derive detailed atomic arrangements in liquid metals.