EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Rational Design of Single-Phase High-Entropy Oxides via Large Language Model Data Mining and Explainable Machine Learning

Arthur da Silva Sousa Santos iD, Elena Stojanovska iD, Antonio Augusto Alves iD, Amauri Jardim de Paula iD, Daniel Zanetti de Florio iD, James Moraes de Almeida iD

Journal of Chemical Information and Modeling · 11.05.2026 · 66 (9) · 5234-5248

DOI: 10.1021/acs.jcim.6c00752 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Chemical Information and Modeling, online erschienen am 25.04.2026. Verfasst von Arthur da Silva Sousa Santos und 5 weiteren Autor:innen. Beteiligte Einrichtung: Center for Engineering, Modeling and Applied Social Sciences, Federal University of ABC (UFABC), Av. dos Estados, 5001, Bangú, Santo André, São Paulo…. Hauptthemen laut MEDLINE: Data Mining, Oxides, Entropy, Machine Learning. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 59 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Conselho Nacional de Desenvolvimento Cient?fico e Tecnol?gico, Funda??o de Amparo ? Pesquisa do Estado de S?o Paulo, Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rational design of high-entropy oxides (HEOs) is currently hindered by the scarcity of structured property data in the scientific literature. In this work, we present an end-to-end materials informatics framework that couples large language model (LLM) data mining with interpretable machine learning to predict single-phase stability in HEOs. We deployed agents based on gpt-oss-120b to extract compositions, phases, and synthesis methods from unstructured scientific abstracts. Combined with regular-expression routines, the LLM-based agent achieved an accuracy of 96% in database generation despite the complexity of the task, including on-the-fly inference of relative cation proportions. Subsequently, a range of machine-learning models was trained in an exploratory multiclass classification setting to distinguish canonical HEO crystal structures using several variants of the primary databases obtained by combining different feature subsets. For this task, an XGBoost classifier achieved an F1-score of 86% in a seven-class classification problem, and the best-performing database variant combined primary and statistical features. This optimal database representation was then used to train a neural-network binary classifier to distinguish perovskite from nonperovskite compositions, achieving 97.9% classification accuracy on the test set, whereas the Goldschmidt tolerance factor reached only 67.3% on the same data. These results indicate that the proposed methodology can support the design of HEO compositions with target properties and substantially outperforms traditional descriptor-based approaches. Furthermore, SHAP (SHapley Additive exPlanations) analysis revealed that high-entropy perovskite phase stability is governed by a critical interplay between geometric factors, such as the sum of cation radii, and electronic descriptors, including Sanderson electronegativity and atomization enthalpy. Overall, these findings demonstrate that LLM-driven data mining can overcome data bottlenecks and enable the discovery of physical design rules for complex, multicomponent ceramics.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Data MiningOxidesEntropyMachine LearningLarge Language ModelsBoosting Machine Learning Algorithms

Substanzen

Oxides

Autor:innen und Institutionen

  • Center for Engineering, Modeling and Applied Social Sciences, Federal University of ABC (UFABC), Av. dos Estados, 5001, Bangú, Santo André, São Paulo 09210-580, Brazil.
  • Independent Researcher, Trento, TN 38122, Italy.
  • Ilum School of Science, Brazilian Center for Research in Energy and Materials (CNPEM), Rua Lauro Vannucci, 1020, Fazenda Santa Cândida, Campinas, São Paulo 13083-970, Brazil.
  • Aeronautics Institute of Technology (ITA), Praça Marechal Eduardo Gomes, 50, Vila das Acacias, São José dos Campos, São Paulo 12228-900, Brazil.
  • Center of Natural and Human Sciences, Federal University of ABC (UFABC), Av. dos Estados, 5001, Bangú, Santo André, São Paulo 09210-580, Brazil.

ORCID-Profile: Arthur da Silva Sousa Santos, Elena Stojanovska, Antonio Augusto Alves, Amauri Jardim de Paula, Daniel Zanetti de Florio, James Moraes de Almeida

Förderung

  • Conselho Nacional de Desenvolvimento Cient?fico e Tecnol?gico
  • Funda??o de Amparo ? Pesquisa do Estado de S?o Paulo
  • Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior
  • Instituto Nacional de Ci?ncia e Tecnologia, Materials Informatics

Zitieren

APA

da Silva Sousa Santos, A., Stojanovska, E., Alves, A. A., de Paula, A. J., de Florio, D. Z., & de Almeida, J. M. (2026). Rational Design of Single-Phase High-Entropy Oxides via Large Language Model Data Mining and Explainable Machine Learning. Journal of Chemical Information and Modeling, 66(9), 5234-5248. https://doi.org/10.1021/acs.jcim.6c00752

Vancouver

da Silva Sousa Santos A, Stojanovska E, Alves AA, de Paula AJ, de Florio DZ, de Almeida JM. Rational Design of Single-Phase High-Entropy Oxides via Large Language Model Data Mining and Explainable Machine Learning. Journal of Chemical Information and Modeling. 2026;66(9):5234-5248. doi:10.1021/acs.jcim.6c00752

BibTeX
@article{daSilvaSousaSantos2026,
  author = {da Silva Sousa Santos, Arthur and Stojanovska, Elena and Alves, Antonio Augusto and de Paula, Amauri Jardim and de Florio, Daniel Zanetti and de Almeida, James Moraes},
  title = {Rational Design of Single-Phase High-Entropy Oxides via Large Language Model Data Mining and Explainable Machine Learning},
  journal = {Journal of Chemical Information and Modeling},
  year = {2026},
  volume = {66},
  number = {9},
  pages = {5234--5248},
  doi = {10.1021/acs.jcim.6c00752}
}
Kostenlos starten →