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

Abstract

Background: Research on green-synthesized nanomaterials (GSNs) for environmental remediation is growing rapidly, yet data remains fragmented in non-interoperable formats. Methods: We present OntoNanoMat, a comprehensive semantic resource consisting of a modular OWL 2 DL ontology and a curated dataset of two illustrative case studies serving as proof-of-concept demonstrations. The data was structured into five thematic modules: Identification, Synthesis, Mechanism, Performance, and Provenance. Results: The dataset is provided in three interoperable formats: CSV, JSON, and Turtle (RDF) and is validated through SHACL shapes to ensure structural integrity and FAIR compliance. Conclusions: OntoNanoMat provides a FAIR-compliant (Findable, Accessible, Interoperable, and Reusable) foundation for future machine learning applications and knowledge graph integration in sustainable nanotechnology.

Dataset: https://doi.org/10.5281/zenodo.18201276

Dataset License: CC-BY 4.0

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.

To provide a clear conceptual overview of the resource’s architecture, the modular design of OntoNanoMat is illustrated in Figure 1. This diagram represents the core semantic engine and its five peripheral thematic modules, highlighting the logical dependencies between synthesis conditions, experimental mechanisms, and remediation performance.

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 [6].

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. Specification of the 35 attributes included in the OntoNanoMat dataset.

Module	Attribute Group	Parameter Name (CSV Header)	Data Type	Example/Unit
I. Identification	Identity	case_id, nanomaterial_id, nanomaterial_name	String	CS1, MB_Adsorbent
	Taxonomy	nanomaterial_type, dataset_identifier	Categorical	Magnetic nanocomposite
II. Synthesis	Biogenic Process	green_synthesized, synthesis_route, precursor_type	Boolean/Str	True, Plant extract
	Sustainability	solvent_type, solvent_greenness, renewable_precursor	String/Bool	Aqueous, Low-toxicity
III. Mechanism	Physical Properties	surface_area_m2_per_g, zeta_potential_mV, band_gap_eV	Float	m2/g, mV, eV
	Conditions	pH, temperature_K, contact_time_min	Float	7.0, 298.0, 120.0
	Photochemistry	irradiance_W_per_m2, wavelength_nm	Float	W/m2, nm
IV. Performance	Efficiency	removal_efficiency_percent, qmax_mg_per_g, rate_constant	Float	%, mg/g, min−1
	Stability	cycles, recyclability_comment, energy_demand_index	Int/String	5, Material reused
V. Provenance	Citation	provenance_publication_title, provenance_publication_doi	String	DOI URL
	Metadata	provenance_publication_year, notes	Integer/Str	2020, Illustrative case

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. Ontology Metrics and Complexity

The OntoNanoMat ontology was developed following a modular design pattern. To provide a quantitative overview of its scope, the key metrics extracted from Protégé 5.5 are summarized in

Table 2. Summary of OntoNanoMat ontology metrics and complexity.

Metric Type	Ontology Element	Count
Axioms	Total number of axioms	462
	Logical axiom count	288
	Declaration axioms	165
Entities	Class count	86
	Object property count	41
	Data property count	38
	Individual count (Proof-of-concept)	2
Reasoning	Description Logic (DL) Expressivity	ALCRIF(D)

.

2.5. 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. Literature Search Strategy and Data Acquisition

A systematic literature search was conducted across Scopus, Web of Science, and Google Scholar to identify peer-reviewed studies on green-synthesized nanomaterials for environmental remediation (2018–2025).

Table 3. Systematic search strategy details.

Database	Full Search String	Results (Records Retrieved)
Scopus/Web of Science	(“green synthesis” OR “biogenic synthesis”) AND (“nanomaterials” OR “nanoparticles”) AND (“environmental remediation” OR “dye removal” OR “heavy metal removal”)	1842
Google Scholar	allintitle: “green synthesis” nanomaterials remediation	320
Total Identified		2162

includes the search string that was applied to titles, abstracts, and keywords.

3.2. Inclusion and Exclusion Criteria

The following criteria were used to select the illustrative case studies:

• Inclusion: (1) Peer-reviewed articles; (2) Full reporting of synthesis precursors and conditions; (3) Quantitative performance data (removal efficiency or $q_{max}$); (4) Characterization data (size, surface area, or zeta potential).
• Exclusion: (1) Review papers or conference abstracts; (2) Studies lacking precise chemical dosage or pH values; (3) Proprietary materials with non-disclosed synthesis routes.

3.3. Ontology Alignment

To ensure semantic interoperability within the nano-EHS (Environmental Health and Safety) domain, OntoNanoMat was aligned with established vocabularies. Mapping was performed using owl:equivalentClass and skos:exactMatch properties to terms in eNanoMapper (ENM) for nanomaterial characterization and NanoCommons for experimental workflow metadata. Provenance was captured using the PROV-O recommendation to ensure traceability of the curated data.

3.4. 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.

3.5. 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.

3.6. 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 (v. 3.9.12) 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.

3.7. 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. Structural Integrity and SHACL Validation

The structural integrity of the OntoNanoMat dataset was verified using SHACL (Shapes Constraint Language). A dedicated shapes file (shacl_shapes.ttl) was developed to enforce constraints on the 35 attributes, ensuring that mandatory fields (e.g., DOI, nanomaterial type) are present and that numerical values (e.g., pH, removal efficiency) fall within physically plausible ranges. The validation was performed using the pyshacl library, achieving a 100% compliance rate for the provided case studies.

4.2. Logical Consistency and Reasoning

The OntoNanoMat ontology was subjected to automated reasoning using the HermiT 1.4.3 reasoner within the Protégé 5.5 environment. The ontology was found to be logically consistent, with no unsatisfiable classes or incoherent property hierarchies detected. This ensures that the Description Logic (DL) expressivity ALCRIF(D) is correctly implemented for downstream semantic applications.

4.3. High-Fidelity Data Validation (Y-Scrambling)

To validate the ‘high-fidelity’ nature of the curated attributes, a target permutation test (Y-scrambling) was conducted (see Table S1 in Supplemental Materials). A Random Forest regressor was trained to predict removal efficiency using the 35 descriptors. While the true model achieved a cross-validated R² of 0.93 ± 0.04, the 200 permuted iterations resulted in a mean R² of −0.02 ± 0.08. The clear separation between the observed performance and the null distribution (empirical p < 0.005) confirms that the dataset captures robust chemical relationships rather than chance correlations.

4.4. Semantic Validation (SHACL)

To ensure the dataset follows the required structural constraints, we applied the 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.

4.5. 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. Software Environment

To interact with the OntoNanoMat resource, the following software versions are recommended: Protégé 5.5 or higher for ontology editing; Python 3.9+ with the rdflib 6.0+ library for RDF processing; and any SHACL-compliant validator such as pyshacl 0.17+ for structural integrity checks.

5.2. Data Access Example

Users can programmatically access the dataset using the provided snippets. The following SPARQL query demonstrates how to retrieve all green-synthesized nanomaterials with a removal efficiency greater than 90%.

# Example: Retrieve all green-synthesized nanomaterials and their removal efficiency    PREFIX onm: <https://zenodo.org/record/18201276/files/ontonanomat.owl#>    SELECT ?material ?efficiency    WHERE {    ?s a onm:GreenNanomaterial;    onm:hasName ?material;    onm:hasRemovalEfficiency ?efficiency.    FILTER(?efficiency > 90.0)    }

5.3. 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.

5.4. 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.

5.5. 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.

6. Conclusions and Future Work

This work presents OntoNanoMat, a novel semantic resource designed to overcome the fragmentation and lack of interoperability in data related to green-synthesized nanomaterials for environmental remediation. By integrating a modular OWL 2 DL ontology with a high-fidelity dataset—validated through SHACL shapes and predictive target permutation tests—this resource provides a robust framework for the FAIR (Findable, Accessible, Interoperable, and Reusable) management of nanotechnology data.

The modular architecture, encompassing Identification, Synthesis, Mechanism, Performance, and Provenance, ensures that both technical and experimental metadata are captured systematically. The two illustrative case studies provided as proof-of-concept demonstrate the utility of the ontology for automated reasoning and precise data retrieval via SPARQL, facilitating the transition from unstructured laboratory reports to machine-actionable knowledge graphs.

Future Work Will Focus on Three Key Strategic Areas

• Scaling and Community Curation: We aim to expand the dataset from the current illustrative cases to a large-scale curated repository by implementing a community-driven ingestion pipeline, allowing researchers to submit biogenic synthesis data directly into the OntoNanoMat schema.
• Ontological Benchmarking: As suggested during the peer-review process, future iterations will include a formal comparative analysis and benchmarking against other emerging nanomaterial ontologies. This will further refine the mapping with eNanoMapper and NanoCommons, ensuring deeper cross-domain harmony.
• Machine Learning Integration: Leveraging the high-fidelity nature of the semantically annotated data, we plan to develop hybrid models that combine semantic reasoning with deep learning to predict the remediation efficiency of new green-synthesis routes, accelerating the discovery of sustainable environmental technologies.

Ultimately, OntoNanoMat serves as a foundational building block for the digital transformation of green nanotechnology, providing the necessary semantic infrastructure to turn heterogeneous experimental results into a collective and interoperable scientific asset.