Capabilities of large language models in interpreting melting and solidification dynamics of metal powders from molecular dynamics data

Powder bed fusion of metallic powders involves complex melting and solidification processes at the atomic scale. In this study, we investigated the capability of ten foundation large language models to automatically interpret molecular dynamics simulation data of titanium nanopowders subjected to rapid thermal cycling. Spatial and structural feature descriptors were extracted as collective variables and provided to the language models in a zero-shot setting. The models correctly identified fundamental physical processes such as surface melting, phase transitions, and recrystallization. However, phenomena requiring complex cross-referencing of multiple indicators remained challenging for most models, with performance closely correlated with overall model capability. As a first step toward automated trajectory interpretation, these findings highlight both the potential and limitations of foundation models in computational materials science.

Authors

Institutions

Publication Details

Journal
Materials Letters
Published
2026-11-16
DOI
https://doi.org/10.1016/j.matlet.2026.141995
Primary Topic
Machine Learning in Materials Science
Type
article
Field-Weighted Citation Impact
0.00

Funders

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

Capabilities of large language models in interpreting melting and solidification dynamics of metal powders from molecular dynamics data

Akihiko Chiba, Yoichi Takagishi
3 links
Materials Letters
Machine Learning in Materials Science
article

Capabilities of large language models in interpreting melting and solidification dynamics of metal powders from molecular dynamics data

Akihiko Chiba, Yoichi Takagishi
article en

Abstract

Powder bed fusion of metallic powders involves complex melting and solidification processes at the atomic scale. In this study, we investigated the capability of ten foundation large language models to automatically interpret molecular dynamics simulation data of titanium nanopowders subjected to rapid thermal cycling. Spatial and structural feature descriptors were extracted as collective variables and provided to the language models in a zero-shot setting. The models correctly identified fundamental physical processes such as surface melting, phase transitions, and recrystallization. However, phenomena requiring complex cross-referencing of multiple indicators remained challenging for most models, with performance closely correlated with overall model capability. As a first step toward automated trajectory interpretation, these findings highlight both the potential and limitations of foundation models in computational materials science.

Materials LettersVol. 429
Tohoku University (JP), New Industry Creation Hatchery Center, Tohoku University (JP)
New Energy and Industrial Technology Development Organization
Machine Learning in Materials Science
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