Towards knowledge-based workflows: a semantic approach to atomistic simulations for mechanical and thermodynamic properties

TL;DR

The paper tackles reproducibility and interoperability challenges in atomistic simulations by introducing semantically annotated workflows and knowledge graphs that capture provenance and align with domain ontologies. It presents a two-stage architecture: a property-calculation stage and a knowledge-graph creation stage, with modular workflow nodes and ontology-driven metadata (CMSO, ASMO, PROV-O) rendered as RDF. A suite of property-calculation workflows—encompassing equation of state, elastic properties, mechanical loading, thermal properties, free energy, nanoindentation, and defect energetics—are demonstrated on iron (Fe) using diverse interatomic potentials, and results are queryable via SPARQL within a knowledge graph. The framework supports cross-method comparisons, Hall-Petch validation, and potential AI-assisted workflow selection, representing a significant step toward FAIR, AI-ready, knowledge-based computational materials science.

Abstract

Mechanical and thermodynamic properties, including the influence of crystal defects, are critical for evaluating materials in engineering applications. Molecular dynamics simulations provide valuable insight into these mechanisms at the atomic scale. However, current practice often relies on fragmented scripts with inconsistent metadata and limited provenance, which hinders reproducibility, interoperability, and reuse. FAIR data principles and workflow-based approaches offer a path to address these limitations. We present reusable atomistic workflows that incorporate metadata annotation aligned with application ontologies, enabling automatic provenance capture and FAIR-compliant data outputs. The workflows cover key mechanical and thermodynamic quantities, including equation of state, elastic tensors, mechanical loading, thermal properties, defect formation energies, and nanoindentation. We demonstrate validation of structure-property relations such as the Hall-Petch effect and show that the workflows can be reused across different interatomic potentials and materials within a coherent semantic framework. The approach provides AI-ready simulation data, supports emerging agentic AI workflows, and establishes a generalizable blueprint for knowledge-based mechanical and thermodynamic simulations.

Figures (7)

• Figure 1: Data journey proposed for atomistic simulations, representing data resulting from atomistic software codes into a knowledge graph.
• Figure 2: Workflow diagram showing the general sequence of steps used in all simulations, from structure creation to validation. Successful results are annotated with ontologies and added to the knowledge graph. The 'KG creation' workflow is described in Section \ref{['sec:semantic']}, while the 'Property calculation' workflows are described in further detail in Section \ref{['sec:workflows']}.
• Figure 3: (a) Conceptual dictionary snippet in JSON format. (b) Schematic representation of the RDF serialization of the metadata fields shown in the JSON. The identifiers are shortened for visualization purposes, DOIs or unique identifiers are assigned to all instances.
• Figure 4: Thermophysical properties of bcc Fe as calculated with the EAM01 potential: (a) Energy-volume curve in comparison with those of different crystal structures in Fe, (b) flow stress with increasing grain size, (c) specific heat, (d) coefficient of thermal expansion, (e) phase transition bcc to hcp as a function of pressure, (f) hardness and forces from nanoindentation.
• Figure 5: Results of SPARQL queries from the knowledge graph: (a) Bulk modulus and (b) elastic constant, $c_{11}$, with different EAM potentials. (c) Formation energy of substitutional impurities (X=Cu, Si, Al, Mg) in Fe computed using the GRACE GRACElysogorskiy2025 model.