CoChat
Sign Up
Cited by 2Metal-Organic Frameworks: Synthesis and ApplicationsMachine Learning in Materials ScienceCarbon Dioxide Capture Technologies

A large language model-based multi-agent methodology for intelligent materials screening: A case study on MOFs for CO2 capture

Open access
Li, Yuanming, Zhao, Jie, Cao, Xiangkun Elvis, Li, Shuangjun·2026·Separation and Purification Technology
DOI 10.1016/j.seppur.2026.137462
Open this paper in CoChat — free account

Abstract

Metal-organic frameworks (MOFs) hold significant potential for CO 2 capture owing to their tunable structures, large surface areas, and versatile chemistries. However, current screening strategies are often limited to specific application scenarios and overlook the intrinsic properties of materials, thereby limiting the transferability of findings. Here, an intelligent evaluation framework was proposed that leverages large language models (LLMs) to integrate semantic reasoning with numerical performance metrics. The approach combines molecular structure, elemental composition, and synthetic feasibility with molecular simulations and process modeling to optimize the design. Adsorption behavior is obtained from atomistic simulations, while cyclic performance is assessed through equilibrium-based process simulations. Notably, energy utilization efficiency is explicitly incorporated as a central performance indicator and assigned higher weighting, thereby emphasizing practical deployment considerations. These numerical indicators are further combined with semantic scores derived from an LLM-based multi-agent system, enabling a balanced, interpretable ranking of candidate MOFs. This dual-level strategy reconciles rigorous optimization with engineering feasibility, safety, and compliance. By aligning numerical robustness with semantic interpretability, the methodology addresses the scenario-dependence of conventional screening methods and provides a scalable pathway for intelligent material evaluation. Beyond CO 2 capture, the framework is readily extensible to diverse adsorption processes and evaluation metrics, highlighting its potential as a next-generation paradigm for material screening and decision support. • Introduces an LLM-based multi-agent framework for MOF screening in CO 2 capture. • Integrates molecular simulations with process modeling for robust evaluation. • Emphasizes energy efficiency as a core metric in material screening. • Dual-level ranking balances numerical rigor with semantic interpretability. • Provides a scalable and extensible pathway for intelligent materials discovery.

Integrity signals

Registry

Crossref · 10.1016/j.seppur.2026.137462

Cites

39 of 39 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

We couldn’t retrieve this paper’s full text, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The remaining literature was distributed across general materials-workflow papers, atomistic simulation infrastructures, MOF and porous-material screening, and polymer simulation.

    “The remaining literature was distributed across general materials-workflow papers, atomistic simulation infrastructures, MOF and porous-material screening, and polymer simulation [17,21,22,24-26,32,35-42].”

    From From Scientific Copilots to Tool-Grounded Autonomy: AI Agents in Simulation-Driven Materials Discovery (2026) · cites this paper as [26]

  • These papers typically split the workflow into role-specific agents for planning, structure retrieval, simulation execution, analysis, and quality control.

    “These papers typically split the workflow into role-specific agents for planning, structure retrieval, simulation execution, analysis, and quality control [18,19,21,25,26,35-38,41,42].”

    From From Scientific Copilots to Tool-Grounded Autonomy: AI Agents in Simulation-Driven Materials Discovery (2026) · cites this paper as [26]

  • In other papers, the workflow could be reconstructed only partially from supporting information, and prompt logs, failure traces, or decision histories remained incomplete.

    “In others, the workflow could be reconstructed only partially from supporting information, and prompt logs, failure traces, or decision histories remained incomplete [17,22,24,26,27,29,33,34].”

    From From Scientific Copilots to Tool-Grounded Autonomy: AI Agents in Simulation-Driven Materials Discovery (2026) · cites this paper as [26]

  • Other studies act more as discovery managers around retrieval, ranking, surrogate models, or protocol navigation.

    “Others act more as discovery managers around retrieval, ranking, surrogate models, or protocol navigation [19,20,26,30,41].”

    From From Scientific Copilots to Tool-Grounded Autonomy: AI Agents in Simulation-Driven Materials Discovery (2026) · cites this paper as [26]

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Li, Y., Zhao, J., Cao, X. E., & Li, S.
Affiliations
  • Korea University (KR)
  • National University of Singapore (SG)
  • Imperial College London (GB)
Corresponding author
  • Cao, Xiangkun Elvis
  • Li, Shuangjun
ORCID
Venue
Separation and Purification Technology
Publisher
Elsevier BV
Published
2026-07
Volume
394
Pages
137462
ISSN
1383-5866 (ISSN-L)
DOI
10.1016/j.seppur.2026.137462
OpenAlex
W7133297633
Cited by
2
References
39
Citations by year
2026: 2
FWCI
2.64
Type
journal article
Language
en
Open access
Licence
Locations
Separation and Purification Technology (publishedVersion, cc-by)
Funders
National University of Singapore
Topics
  • Physical Sciences › Chemistry › Inorganic Chemistry › Metal-Organic Frameworks: Synthesis and Applications
  • Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
  • Physical Sciences › Engineering › Mechanical Engineering › Carbon Dioxide Capture Technologies
Keywords
metal-organic frameworks, CO2 capture, materials screening, large language models, multi-agent systems, semantic reasoning, adsorption performance, energy use efficiency, process simulation, candidate ranking, materials discovery

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review