Building Atomic Structures from Natural Language
Abstract
Large language models can write the input parameters of a materials simulation from a prompt, but they cannot build the atomic structure those parameters describe. A crystal, a surface slab or a mismatched interface must satisfy global constraints-periodicity along specific axes, a given vacuum thickness, a commensurate registry, sensible bond lengths. Generation one token at a time enforces none of them, and a violation is invisible in the output. We present ATLAS, a framework that makes structural construction scriptable. Seven components-analyzer, LLM skill, JSON format, build engine, validator, scorer and web interface-are organized around a component algebra of named structures and spaces. Every specification carries a check list, so a build is verified against the request rather than merely parsed from it, and builds are scored against the user's intent. Requests state physical sizes, and the engine computes the replication that reaches them. The translation is the one stochastic stage, and it is recorded, so a run replays with the model held fixed-which is what makes the framework's own behaviour measurable rather than merely observed. Fourteen structural classes are demonstrated, from bulk cells and nanoparticles through surfaces, defects and interfaces to amorphous networks and liquids. We outline how the same loop generates training sets for machine-learning force fields.
Related
- Crystalline ASR products: Review of mineral candidates and molecular simulations of their atomic structures, thermodynamic data and thermoelastic properties
- EpNet-Si: Atomic Form Factors Learned from Band Structures
- MASON: Materials Atomistic Structure builder Operated in Natural language
- Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures
- Scaling Hindi Quantum Natural Language Processing through Automatic Pregroup Supertagging
BibTeX
@misc{li2026building,
title = {{Building Atomic Structures from Natural Language}},
author = {Pai Li},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.29595},
eprint = {2609.29595},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.29595},
}
RIS
TY - UNPB
TI - Building Atomic Structures from Natural Language
AU - Pai Li
PY - 2026
DA - 2026/09/25
PB - arXiv
DO - 10.48550/arxiv.2609.29595
UR - https://arxiv.org/abs/2609.29595
AB - Large language models can write the input parameters of a materials simulation from a prompt, but they cannot build the atomic structure those parameters describe. A crystal, a surface slab or a mismatched interface must satisfy global constraints-periodicity along specific axes, a given vacuum thickness, a commensurate registry, sensible bond lengths. Generation one token at a time enforces none of them, and a violation is invisible in the output. We present ATLAS, a framework that makes structural construction scriptable. Seven components-analyzer, LLM skill, JSON format, build engine, validator, scorer and web interface-are organized around a component algebra of named structures and spaces. Every specification carries a check list, so a build is verified against the request rather than merely parsed from it, and builds are scored against the user's intent. Requests state physical sizes, and the engine computes the replication that reaches them. The translation is the one stochastic stage, and it is recorded, so a run replays with the model held fixed-which is what makes the framework's own behaviour measurable rather than merely observed. Fourteen structural classes are demonstrated, from bulk cells and nanoparticles through surfaces, defects and interfaces to amorphous networks and liquids. We outline how the same loop generates training sets for machine-learning force fields.
ER -