Partial radial distribution functions for a two-component glassy solid, GeSe , from scattering experimental data...

Area of Science:

• Materials Science
• Computational Physics
• Machine Learning

Background:

• The Hopfield neural network (HNN) is effective for inverse problems in monoatomic liquids.
• Retrieving structural functions like g(r) and C(r) from scattering data is crucial for materials characterization.

Purpose of the Study:

• To extend the HNN method for analyzing complex two-component glassy solids, specifically GeSe3.
• To validate the HNN's capability in accurately determining structural properties of more intricate materials.

Main Methods:

• Applied the Hopfield neural network to solve coupled equations for a GeSe3 glassy solid.
• Iteratively adjusted initial conditions based on previous run results to refine calculations.
• Utilized experimental scattering data as input for the network.

Main Results:

• Successfully retrieved the radial distribution function (g(r)) for the complex GeSe3 system.
• Achieved accurate peak intensities and large-r behavior by refining calculations.
• Demonstrated the HNN's robustness in handling more complex material structures.

Conclusions:

• The Hopfield neural network method is a robust tool for analyzing the structure of complex glassy solids.
• The iterative adjustment of initial conditions enhances accuracy for intricate systems.
• This approach advances the application of neural networks in materials structure determination from experimental data.