Searcharxiv⌕ Search

arXiv subjects

Qiyao He

Publications and source records attributed to Qiyao He.

4 recordsLinked to original sources

Gel-Chemistry-Controlled Heavy-Metal Transport and Retention in Cementitious Gel Nanopores: A Molecular Dynamics Study

Cementitious materials are widely used for hazardous-waste encapsulation, yet molecular mechanisms governing heavy-metal ion retention across gel chemistries remain insufficiently resolved. Previous MD studies have focused on individual gel systems and pore-averaged metrics, limiting direct comparison across gel chemistries. Here, classical molecular dynamics simulations were used to investigate adsorption-controlled mobility of Pb2+, Ba2+, and Cs+ in nanopores of C-S-H, C-(N)-A-S-H, and N-A-S-H gels. Combining pore-averaged and spatially resolved diffusivity, residence-time analysis, ion-density profiles, 2D adsorption maps, radial distribution functions, coordination analysis, and interfacial binding-strength descriptors establishes a framework linking gel surface chemistry to mobility suppression while distinguishing local mobility from spatial ion partitioning effects on pore-averaged transport. Ion mobility is reduced in all gel nanopores relative to bulk, but the extent and mechanism vary strongly with gel chemistry. C-(N)-A-S-H with higher interfacial Al/Si substitution exhibits the strongest retention, driven by ion accumulation around Al-linked oxygen species via an ion-exchange-like mechanism with charge-balancing Na+. C-S-H immobilizes ions primarily through surface hydroxyl oxygens and Ca-mediated linkages, whereas N-A-S-H exhibits more distributed binding environments. Pb2+ and Ba2+ show broadly similar immobilization mechanisms, whereas Cs+ exhibits more gel-dependent interactions with silicate and aluminosilicate oxygen sites. A relative total binding strength (rTBS) descriptor shows a strong positive correlation with ion immobilization across gel types, ion species, and pore sizes. These results clarify gel- and ion-specific retention mechanisms and link interfacial binding strength to pore-averaged mobility suppression in idealized cementitious nanopores.

physics.chem-ph↗

Large language model-enabled automated data extraction for concrete materials informatics

The promise of data-driven materials discovery remains constrained by the scarcity of large, high-quality, and accessible experimental datasets. Here, we introduce a generalizable large language model (LLM)-powered pipeline for automated extraction and structuring of materials data from unstructured scientific literature, using concrete materials as a representative and particularly challenging example. The pipeline exhibits robust performance across a broad range of LLMs and achieves an $F_1$ score of up to 0.98 for diverse composition--process--property attributes. Within one hour, it extracts nearly 9,000 high-quality records with over 100 attributes from a corpus screened from more than 27,000 publications, enabling the construction of the largest open laboratory database for blended cement concrete. Machine learning analyses underscore the importance of large, diverse, and information-rich datasets for enhancing both in-distribution accuracy and out-of-distribution generalization to unseen materials. The proposed pipeline is readily adaptable to other materials domains and accelerates the development of scalable data infrastructures for materials informatics.

cond-mat.mtrl-sci↗

Automating literature screening and knowledge synthesis with large language models

Scientific literature is expanding rapidly, yet much of its knowledge remains difficult to organize and reuse due to the substantial workload needed to navigate large, heterogeneous literature corpora. Here, we develop a generalizable large language model (LLM)-powered framework that automates literature screening, structured information extraction, and knowledge synthesis within a unified workflow. We benchmarked 11 state-of-the-art proprietary and open-weight LLMs across three materials science screening tasks differing in scope, corpus size, and class imbalance. Seven LLMs achieved balanced accuracies above 0.90 across all tasks, with the best-performing LLM reaching 0.96--0.98, demonstrating the potential of LLMs to accurately screen large, highly imbalanced scientific corpora. The framework also substantially reduces manual effort, screening 10,000 records within four minutes compared with approximately 83 person-hours for manual screening. Beyond literature screening, the framework extracts structured information from full-text articles, enabling content-level quantification of research landscapes that extends traditional bibliometric analysis based on publication metadata. This capability reveals methodological trends and underexplored directions, as shown for machine learning in concrete materials science. We further demonstrate its utility for information and data discovery by identifying publications containing silicate glass density data that are difficult to retrieve using conventional search strategies, which can support downstream hypothesis development, research design, and data-driven scientific discovery. Overall, these results demonstrate the broad applicability of LLM-powered literature analysis across scientific domains and its potential to transform large, unstructured literature corpora into structured, analyzable knowledge at scale.

cs.CL↗

Reactivity-Informed Machine Learning for Performance Prediction and Design Space Exploration of Alkali-Activated Slag

Establishing quantitative relationships among mix design, raw material properties, curing conditions, and performance remains a long-standing challenge in cementitious materials, particularly for alkali-activated materials with variable precursor and activator chemistry. Here, we curated the largest literature-derived alkali-activated slag (AAS) dataset to date, comprising over 3100 compressive strength records, 155 chemically distinct ground granulated blast-furnace slags (GGBSs), and 24 attributes incorporating precursor chemistry, fineness, and reactivity. Multiple machine learning (ML) algorithms were benchmarked across progressively enriched feature scenarios, demonstrating that integrating GGBS compositions, fineness, curing conditions, and specimen geometry improves predictive performance. The average metal oxide dissociation energy (AMODE), a physically interpretable representation of precursor reactivity, provides a compact alternative descriptor to explicit oxide compositions while enabling comparable predictive performance. Model interpretation revealed physically consistent trends from heterogeneous data, including non-monotonic effects of Na2O dosage and silicate modulus, reduced predicted strength at higher water content and larger specimen size, and coupled oxide-level effects more coherently represented by AMODE than by individual oxide contents. Statistically constrained design space exploration reveals reactivity-dependent trade-offs among strength, embodied CO2 emissions, and cost. The design maps identify high-strength regions with substantially lower CO2 emissions than OPC-based references at similar cost. Overall, this work demonstrates how reactivity-informed ML can extract physically meaningful trends from heterogeneous AAS data and guide source-dependent binder design. The curated dataset is publicly accessible to support advances in cement and concrete research.

cond-mat.mtrl-sci↗