CoSP: Reconfigurable Multi-State Metamaterial Inverse Design via Contrastive Pretrained Large Language Model

2025-11-20

AI Summary

Current deep learning–based inverse design methods struggle to jointly optimize the multi-state, multi-band optical responses of reconfigurable multimodal metamaterials (RMMs). To address this, we propose CoSP—a novel framework that, for the first time, couples a contrastively pretrained spectral encoder with a large language model (LLM), endowing the system with electromagnetic physical awareness and natural language comprehension. CoSP employs physics-guided spectral-structural mapping to enable end-to-end inverse generation: given target multi-state optical responses (e.g., reflection/transmission spectra across multiple bands and configurations), it directly synthesizes tunable metasurface structures. Experiments demonstrate that CoSP accurately generates thin-film metasurfaces satisfying arbitrarily specified multi-state, multi-band optical requirements. It significantly outperforms existing approaches in design efficiency, flexibility, and physical consistency, establishing a new paradigm for multifunctional, reconfigurable photonic devices.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Multi-modal Vision
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Metamaterials, known for their ability to manipulate light at subwavelength scales, face significant design challenges due to their complex and sophisticated structures. Consequently, deep learning has emerged as a powerful tool to streamline their design process. Reconfigurable multi-state metamaterials (RMMs) with adjustable parameters can switch their optical characteristics between different states upon external stimulation, leading to numerous applications. However, existing deep learning-based inverse design methods fall short in considering reconfigurability with multi-state switching. To address this challenge, we propose CoSP, an intelligent inverse design method based on contrastive pretrained large language model (LLM). By performing contrastive pretraining on multi-state spectrum, a well-trained spectrum encoder capable of understanding the spectrum is obtained, and it subsequently interacts with a pretrained LLM. This approach allows the model to preserve its linguistic capabilities while also comprehending Maxwell's Equations, enabling it to describe material structures with target optical properties in natural language. Our experiments demonstrate that CoSP can design corresponding thin-film metamaterial structures for arbitrary multi-state, multi-band optical responses, showing great potentials in the intelligent design of RMMs for versatile applications.

Problem

• Designing reconfigurable metamaterials with multi-state optical switching capabilities
• Overcoming limitations of existing deep learning methods in inverse design
• Enabling natural language description of structures with target optical properties

Innovation

• Contrastive pretraining on multi-state spectrum data
• Integration of spectrum encoder with large language model
• Natural language description of target optical properties