Artificial intelligence guided search for van der Waals materials with high optical anisotropy

Area of Science:

• Materials Science
• Photonics
• Computational Chemistry

Background:

• Van der Waals (vdW) materials possess unique optical properties, particularly optical anisotropy, crucial for advanced photonic applications.
• Manual screening of vdW materials for giant optical anisotropy is time-consuming and inefficient for rapid discovery.

Purpose of the Study:

• To develop and apply machine learning (ML) and geometrical approaches for efficient identification of vdW materials with giant optical anisotropy.
• To accelerate the discovery and optimization of novel vdW materials for photonic applications.

Main Methods:

• Utilized geometrical clustering based on birefringence and crystallographic parameters.
• Employed deep learning architectures, including the Atomistic Line Graph Neural Network, for predictive modeling.
• Validated ML predictions using density functional theory and ellipsometry measurements.

Main Results:

• Developed an accurate ML model capable of predicting optical anisotropy in vdW materials.
• Geometrical approach provided insights into structure-property relationships.
• Experimental verification confirmed the model's predictions for 2H-MoTe₂ and CdPS₃.

Conclusions:

• Machine learning significantly streamlines the search for vdW materials with exceptional optical anisotropy.
• This approach holds great potential for discovering and optimizing materials for next-generation photonic devices.