Disorder by design: A data-driven approach to amorphous semiconductors without total-energy functionals

Area of Science:

• Materials Science
• Condensed Matter Physics
• Computational Materials Science

Background:

• Determining the 3D structure of amorphous semiconductors from diffraction data is a complex inverse problem.
• Traditional methods often require detailed knowledge of local chemistry (e.g., energy functionals), which is challenging for amorphous materials.

Purpose of the Study:

• To develop and validate a method for reconstructing realistic 3D models of tetrahedral amorphous semiconductors using X-ray diffraction data.
• To solve the inverse problem without relying on total-energy calculations or force fields.

Main Methods:

• Formulating the material-structure determination as a multi-objective optimization problem.
• Utilizing X-ray diffraction data combined with Monte Carlo methods.
• Employing minimal structural constraints for model reconstruction.

Main Results:

• Generated highly realistic 3D models of amorphous silicon (a-Si) with very few coordination defects (≤1%).
• Achieved a narrow bond-angle distribution (9-11.5°) and a realistic electronic band gap (0.8-1.4 eV).
• Data-driven models accurately reproduced experimental electronic and vibrational properties, comparable to established models like Wooten-Winer-Weaire.

Conclusions:

• The multi-objective optimization approach is effective for structural determination of complex amorphous materials.
• This data-driven method provides a unique and accurate model of amorphous silicon from diffraction data, resolving previous ambiguities.