Deep Learning-Assisted Fourier Analysis for High-Efficiency Structural Design: A Case Study on Three-Dimensional...

Area of Science:

• Materials Science
• Crystallography
• Computational Chemistry

Background:

• Designing functional materials with specific properties is complex due to the vast number of possible structures.
• Identifying correlations between material structure and properties remains a significant challenge.

Purpose of the Study:

• To develop a universal method for the design and property optimization of periodic structures.
• To enable efficient exploration of parameter spaces for identifying ideal material geometries.

Main Methods:

• A deep-learning-assisted inverse Fourier transform approach for generating arbitrary geometries within crystallographic space groups.
• Application to the design and analysis of three-dimensional (3D) photonic structures.

Main Results:

• The method successfully modeled numerous structures and identified photonic bandgaps within hours.
• Confirmed that network morphologies, particularly the single diamond (dia net), yield the widest photonic bandgaps.
• Discovered a rare lcs topology with superior photonic properties.

Conclusions:

• The Fourier-based method is highly efficient and effective for materials design.
• This approach advances the discovery of next-generation functional materials.
• Demonstrates broad applicability in exploring material design possibilities.