[3C01-05] Data-Driven Surface Science 

10:00 AM - 10:30 AM JST(1:00 AM - 1:30 AM UTC)

[3C03【依頼講演】] Automated design of two-dimensional magnetic thin film patterns for magnonics applications using an agentic design approach

*Ryunosuke Nagaoka¹, Masato Kotsugi¹ (1. Tokyo University of Science (Japan))

Magnonic crystals (MCs) provide magnetic platforms for controlling spin-wave (magnon) band dispersions, but their complex structure–property relationships make rational design challenging. Here, we develop an autonomous design agent that optimizes two-dimensional patterns of magnetic thin films by integrating genetic-algorithm-based topology optimization with frequency-domain micromagnetic simulations. The agent explores a vast combinatorial design space and discovers previously unreported MC structures with bandgaps up to 830% larger than the previous best-performing design, as validated by time-domain simulations. Data-driven visualization further reveals increasingly nonconvex design landscapes for higher-order bands. This agent-based framework enables autonomous discovery of high-performance MCs and provides insights toward general design rules for future magnonic devices.

Password required to view