OntoNanoMat: A Semantic Dataset and Ontology for Green-Synthesized Nanomaterials in Environmental Remediation

1. Summary

The rapid expansion of green nanotechnology has led to a vast but fragmented body of literature regarding the use of green-synthesized nanomaterials (GSNs) for environmental remediation [1]. While these materials offer sustainable alternatives to traditional chemical synthesis, the lack of standardized data structures makes it difficult to perform cross-study comparisons, assess “greenness” objectively, or reuse data for large-scale meta-analyses [2]. To address these challenges, we developed OntoNanoMat, a semantic resource designed to externalize and formalize knowledge in the GSN domain.

The OntoNanoMat dataset was collected through a systematic review of recent scientific literature focusing on green synthesis routes (e.g., biogenic reagents, plant extracts) and their application in removing contaminants like organic dyes and heavy metals from water [3]. The dataset was structured using a modular OWL 2 DL ontology, ensuring that every data point is linked to its chemical precursors, synthesis conditions, and performance indicators (such as removal efficiency and recyclability) [4].

This dataset is a core component of a broader research effort at the University of Guadalajara to digitalize material knowledge and promote FAIR (Findable, Accessible, Interoperable, and Reusable) principles in nanotechnology. The creation of this resource was motivated by the need to provide a “ground truth” for semantic integration frameworks, such as the one described in our corresponding research article [5], where we demonstrate how this ontology-based approach enables the interoperable assessment of material sustainability and performance.

By releasing this dataset in multiple formats (CSV, JSON, and Turtle), we aim to provide a ready-to-use resource for the scientific community. Potential benefits include the facilitation of automated data discovery, the training of machine learning models for predicting nanomaterial efficiency, and the integration of GSN data into global Knowledge Graphs for sustainable chemistry.

2. Data Description

The OntoNanoMat dataset is a curated collection of case studies focusing on the environmental application of green-synthesized nanomaterials. The resource is structured to provide high interoperability between tabular processing, web development, and semantic reasoning. All data files are available in the Zenodo repository https://doi.org/10.5281/zenodo.18201276.

2.1. Dataset Files and Formats

The dataset is distributed in three main formats to support different use cases:

• dataset_case_studies.csv: A UTF-8 encoded tabular file containing the primary data for statistical analysis.
• dataset_case_studies.json: A machine-readable JSON array, ideal for integration into web platforms or NoSQL databases.
• dataset_case_studies.ttl: An RDF serialization in Turtle format. This file links the data instances to the classes and properties defined in the green_nanomaterials_ontology.ttl file.

2.2. Tabular Data Structure

The CSV file consists of 35 columns (attributes) per entry. Table 1 describes the main fields and their interpretation.

2.3. Semantic Mapping and Interpretation

The data is designed to be interpreted through the lens of the OntoNanoMat Ontology. In the Turtle (.ttl) version, each record is transformed into an individual of the class gsn:Nanomaterial.

• Logic Links: The synthesis descriptors are mapped to the gsn:SustainabilityProfile class, while performance metrics are linked to gsn:PerformanceIndicator.
• Units: All numerical values follow standard units: Temperature in Kelvin (K), concentration in g/L, and adsorption capacity in mg/g, as defined by the ontology’s datatype properties.

2.4. Validation Resource

In addition to the dataset, the repository includes green_nanomaterials_queries.rq, a library of SPARQL queries. These queries serve as an “executable documentation” that demonstrates how to retrieve and filter data based on multi-dimensional criteria (e.g., finding materials with high efficiency that also use renewable solvents).

3. Methods

The development of the OntoNanoMat resource followed a three-stage methodology: data acquisition through systematic curation, semantic modeling (ontology design), and data transformation (serialization).

3.1. Data Acquisition and Curation

The case studies included in the dataset were retrieved through a systematic search of peer-reviewed literature published between 2018 and 2025. Search queries were conducted in databases such as Scopus, Web of Science, and Google Scholar using combinations of keywords including “green synthesis”, “nanomaterials”, “environmental remediation”, and “sustainable nanotechnology”.

Data extraction was performed manually to ensure the high fidelity of technical parameters. For each case study, we recorded:

• Synthesis parameters: Solvent type, precursors, and energy indicators.
• Experimental conditions: pH, temperature, and dosage.
• Performance metrics: Removal efficiency and adsorption capacity (q_max).

Numerical values were normalized to standard units (e.g., converting all temperatures to Kelvin and concentrations to mg/L) to facilitate interoperability and comparison.

3.2. Ontology Development

The OntoNanoMat Ontology was developed using the OWL 2 DL (Web Ontology Language) standard. The modeling process followed an iterative approach using Protégé 5.6.x.

• Modularity: The ontology was organized into five core modules (Material, Synthesis, Process, Performance, and Provenance) to allow for independent updates.
• Reusability: Where possible, classes and properties were aligned with existing vocabularies such as PROV-O for provenance and CHEO or ENM for chemical entities.
• Axiomatization: Logical restrictions (SubClassOf and EquivalentTo) were implemented to enable automatic classification of “green-synthesized” materials based on their sustainability profiles.

3.3. Data Transformation and RDFization

To generate the multi-format dataset, we followed these steps:

• Tabular Structuring: The curated data was first organized into a master CSV file.
• Semantic Mapping: Using a custom Python-based mapping script, each CSV row was transformed into an RDF individual (instance).
• Serialization: The data was exported into JSON for web accessibility and Turtle (.ttl) for semantic reasoning. The Turtle version explicitly uses the gsn: namespace defined in the ontology to ensure that the instances are logically bound to their semantic definitions.

3.4. Technical Validation Setup

Validation was not limited to syntax checking. We developed a set of SHACL (Shapes Constraint Language) files to enforce data integrity constraints (e.g., ensuring that any material labeled as “adsorbent” must have an associated q_max value). Finally, a library of SPARQL queries was created to verify that the graph could answer complex competency questions regarding material performance and greenness.

4. Technical Validation

The technical quality and integrity of the OntoNanoMat resource were evaluated through a multi-layered validation pipeline.

4.1. Syntactic and Structural Validation

All RDF serializations in Turtle (.ttl) format were validated using the Apache Jena RIOT tool to ensure compliance with W3C standards. This step confirmed that the data is free of syntax errors and ready for ingestion by any standard triplestore. The JSON and CSV files were also checked for schema consistency and UTF-8 encoding integrity.

4.2. Logical Consistency and Reasoning

The green_nanomaterials_ontology.ttl was subjected to automated reasoning using the HermiT 1.4.3 reasoner within the Protégé environment. No logical inconsistencies or unsatisfiable classes were detected. The hierarchy correctly infers individuals into their respective subclasses (e.g., a material with a “biogenic precursor” property is correctly classified as a gsn:GreenSynthesizedMaterial).

4.3. Semantic Validation (SHACL)

To ensure the dataset follows the required structural constraints, we applied Shapes Constraint Language (SHACL). The validation shapes (provided in the repository) verify that:

• Each Nanomaterial entry is linked to at least one RemediationMechanism.
• Quantitative indicators (like removal_efficiency_percent) are restricted to numerical ranges (0–100).
• Mandatory provenance metadata (DOI and year) is present for every record.

4.4. Competency Question Testing

A library of eight SPARQL queries was used to validate the functional utility of the data. These queries successfully retrieved complex cross-referenced information, such as identifying nanomaterials that achieve >90% efficiency while maintaining a “low-toxicity” solvent profile. This confirms that the resource can answer the domain-specific questions for which it was designed.

5. Usage Notes (or User Notes)

The OntoNanoMat resource is designed for researchers in nanotechnology, environmental science, and data engineering.

5.1. Accessing and Exploring the Data

The dataset and ontology can be accessed via the GitHub repository (for version control and issue tracking) or the Zenodo archive (for the stable, citable version).

• For Nanotechnologists: The dataset_case_studies.csv file can be opened in any spreadsheet software (Excel, Google Sheets) or R/Python environments for quick benchmarking.
• For Knowledge Engineers: The .ttl file should be loaded into Protégé or a triplestore like Apache Jena Fuseki or GraphDB. Users can then execute the provided SPARQL queries to filter materials by specific green chemistry or performance criteria.

5.2. Integration and Extensibility

The modular nature of the ontology allows it to be easily extended. Researchers can add new remediation mechanisms (e.g., membrane filtration) or additional nanomaterial characterization parameters by defining them as subclasses of the existing core classes.

5.3. Software Requirements

No specialized software is required to view the primary data (CSV). However, to fully leverage the semantic features:

• Protégé (v5.5 or higher) is recommended for ontology visualization.
• Python (rdflib library) is suggested for those wishing to programmatically integrate this dataset into machine learning pipelines or larger Knowledge Graphs.