From High-Throughput Screening to Generative Design: Artificial Intelligence-Driven Dielectric Materials Discovery.

Area of Science:

• Materials Science
• Artificial Intelligence
• Condensed Matter Physics

Background:

• High-performance dielectric materials are crucial for advanced electronics, but their design faces trade-offs between breakdown strength and permittivity.
• Current screening methods are limited to existing materials and cannot address complex multiphysics coupling challenges.

Purpose of the Study:

• To review the paradigm shift in dielectric material research from empirical screening to AI-driven autonomous discovery.
• To provide a framework for selecting AI tools for targeted dielectric design.

Main Methods:

• Examining the physical origins of dielectric performance trade-offs and data scarcity issues.
• Surveying physics-informed descriptors, surrogate models, and multiobjective optimization frameworks.
• Analyzing generative architectures (VAEs, GANs, diffusion models) for inverse design.
• Discussing the integration of autonomous laboratories and active learning for closed-loop discovery.

Main Results:

• AI enables inverse design of novel dielectric polymers and inorganic crystals beyond current chemical spaces.
• Physics-informed AI approaches bridge atomic-scale structures and macroscopic dielectric properties.
• Integration of AI with experimental validation accelerates materials development.

Conclusions:

• AI-driven autonomous discovery represents a new era for rational and accelerated dielectric material innovation.
• Addressing barriers in physics embedding, explainability, and synthesizability is key for future progress.
• LLM-driven digital scientists offer a roadmap for next-generation dielectric innovation.