Nanophotonic device design based on large language models: multilayer and metasurface examples.

Area of Science:

• Nanophotonics
• Artificial Intelligence
• Computational Science

Background:

• Large language models (LLMs) excel in language tasks and are expanding into scientific applications.
• The use of LLMs for nanophotonic device design is an emerging and underexplored area.
• Nanophotonic design traditionally requires specialized domain expertise.

Purpose of the Study:

• To investigate the efficacy of LLMs in addressing nanophotonic design challenges.
• To determine if LLMs can enable nonexpert users to design nanophotonic devices.
• To explore LLM capabilities in optical response calculation and inverse design for nanophotonic structures.

Main Methods:

• Utilizing LLMs with in-context learning for numerical simulations of optical responses in multilayer films.
• Employing conversational interaction and feedback loops between LLMs and users for design optimization.
• Fine-tuning LLMs with text-based representations of optical metasurface structures and properties.
• Implementing text-based input/output reversal for generative metasurface design.

Main Results:

• LLMs with in-context learning allow nonexpert users to simulate optical responses of multilayer films.
• Conversational LLM interaction facilitates the creation of optimal multilayer film designs for target optical properties.
• Fine-tuned LLMs can successfully generate metasurface designs tailored to specific properties.

Conclusions:

• LLMs show significant potential to simplify and accelerate nanophotonic design.
• LLMs can democratize nanophotonic design, making it accessible to users without deep domain knowledge.
• This work paves the way for AI-driven innovation in nanophotonics.