VortexChat: Autonomous design of photonic devices from natural language

Abstract The advancement of modern photonics is frequently bottlenecked by device design workflows that rely heavily on manual simulation and expert intuition. Although existing photonic inverse-design methods have largely automated parameter optimization, they still rely on researchers to formulate optimization objectives, construct predefined workflows, and diagnose or recover from unsuccessful optimization trajectories. To address these issues, we introduce VortexChat, an agentic framework that translates natural-language specifications into photonic designs. VortexChat autonomously orchestrates generative models, gradient-based optimization, and automated full-wave electromagnetic simulations, updating design strategies based on feedback. We demonstrate VortexChat through the autonomous design of a broadband terahertz perfect vortex beam multiplexer, a challenging task for conventional manual design approaches. Evaluated on the Vortex100 Benchmark, VortexChat achieves a 71% overall success rate under the simultaneous constraints of efficiency, mode purity, and broadband performance. Experimental measurements validate the resulting device, demonstrating mode purity exceeding 84% across the operating band, together with high efficiency and low inter-channel crosstalk. By reducing reliance on expert intervention, VortexChat offers a route toward more accessible and autonomous development of complex photonic devices.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11591735/v1
Primary Topic
Metamaterials and Metasurfaces Applications
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

VortexChat: Autonomous design of photonic devices from natural language

Song Hee Han, Hongsheng Chen, Shilong Li, Hongsheng Chen et al.
4 links
Research Square
Metamaterials and Metasurfaces Applications
preprint

VortexChat: Autonomous design of photonic devices from natural language

Song Hee Han, Hongsheng Chen, Shilong Li, Hongsheng Chen, Faqian Chong, Yulun Wu
preprint en

Abstract

Abstract The advancement of modern photonics is frequently bottlenecked by device design workflows that rely heavily on manual simulation and expert intuition. Although existing photonic inverse-design methods have largely automated parameter optimization, they still rely on researchers to formulate optimization objectives, construct predefined workflows, and diagnose or recover from unsuccessful optimization trajectories. To address these issues, we introduce VortexChat, an agentic framework that translates natural-language specifications into photonic designs. VortexChat autonomously orchestrates generative models, gradient-based optimization, and automated full-wave electromagnetic simulations, updating design strategies based on feedback. We demonstrate VortexChat through the autonomous design of a broadband terahertz perfect vortex beam multiplexer, a challenging task for conventional manual design approaches. Evaluated on the Vortex100 Benchmark, VortexChat achieves a 71% overall success rate under the simultaneous constraints of efficiency, mode purity, and broadband performance. Experimental measurements validate the resulting device, demonstrating mode purity exceeding 84% across the operating band, together with high efficiency and low inter-channel crosstalk. By reducing reliance on expert intervention, VortexChat offers a route toward more accessible and autonomous development of complex photonic devices.

Research Square
University of the Witwatersrand (ZA), Zhejiang University (CN)
National Natural Science Foundation of China, Young Scientists Fund
Metamaterials and Metasurfaces Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon