Crystal-structure design by agentic AI in a language of motifs
Abstract
Data-driven materials discovery interpolates more reliably than it extrapolates and seldom reveals new structure types. We present MatEvolve, an agentic-AI framework that designs crystals in an interpretable \emph{language of motifs}, in which each crystal is written as a \emph{motif profile} that describes the recurring geometric patterns---the \emph{motifs}---composing the crystal. The motif profile serves not merely as a description of a material but as the medium for material design: the agent edits the profile with a stated rationale and constructs a crystal from the modified one; the most promising candidates are then validated by first-principles calculations. MatEvolve is built on the state-of-the-art language model Claude Fable~5 without fine-tuning. It is applied to the design of rare-earth-lean permanent-magnet compounds and finds novel (meta)stable crystals that are both new structural prototypes and magnet compounds; under an equal validation budget, it does so an order of magnitude more often than generative models. Beyond design, analyzing the human-readable profiles of the discovered crystals reveals their structure--property relationships.
Related
- Automated mode-aware design of nonlinear photonic crystal cavities for intrinsic second harmonic generation
- Agentic inverse design of spatially programmed materials
- Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures
- LATHE: LAnguage-driven Toolkit for Hypothesis-based crystal Editing
- Crystal structure of URu4Si2, a new Ru-rich phase in the U–Ru–Si ternary system
BibTeX
@misc{le2026crystalstructure,
title = {{Crystal-structure design by agentic AI in a language of motifs}},
author = {Dinh-Khiet Le and Minh-Quyet Ha and Hong-Phuc Vu-Dinh and Viet-Cuong Nguyen and Huan Tran and Truyen Tran and Takashi Miyake and Hiori Kino and Hieu-Chi Dam},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2608.15900},
eprint = {2608.15900},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.15900},
}
RIS
TY - UNPB
TI - Crystal-structure design by agentic AI in a language of motifs
AU - Dinh-Khiet Le
AU - Minh-Quyet Ha
AU - Hong-Phuc Vu-Dinh
AU - Viet-Cuong Nguyen
AU - Huan Tran
AU - Truyen Tran
AU - Takashi Miyake
AU - Hiori Kino
AU - Hieu-Chi Dam
PY - 2026
DA - 2026/10/21
PB - arXiv
DO - 10.48550/arxiv.2608.15900
UR - https://arxiv.org/abs/2608.15900
AB - Data-driven materials discovery interpolates more reliably than it extrapolates and seldom reveals new structure types. We present MatEvolve, an agentic-AI framework that designs crystals in an interpretable \emph{language of motifs}, in which each crystal is written as a \emph{motif profile} that describes the recurring geometric patterns---the \emph{motifs}---composing the crystal. The motif profile serves not merely as a description of a material but as the medium for material design: the agent edits the profile with a stated rationale and constructs a crystal from the modified one; the most promising candidates are then validated by first-principles calculations. MatEvolve is built on the state-of-the-art language model Claude Fable~5 without fine-tuning. It is applied to the design of rare-earth-lean permanent-magnet compounds and finds novel (meta)stable crystals that are both new structural prototypes and magnet compounds; under an equal validation budget, it does so an order of magnitude more often than generative models. Beyond design, analyzing the human-readable profiles of the discovered crystals reveals their structure--property relationships.
ER -