Semantic resources for FAIR description of computational workflows
Practical use of domain ontologies, controlled vocabularies, provenance models and standardised units, illustrated by the DFT → MLIP → MD workflow in materials science.
Consistent terminology makes it possible to describe materials, computational methods, parameters, results and their provenance so that these descriptions can be interpreted unambiguously by both people and machine-based systems.
The Competence Center recommends reusing established international vocabularies, ontologies and persistent identifiers for concepts whenever possible. Local terms should be introduced only for those workflow elements for which no suitable external semantic resource is available.
Reuse existing semantics before creating new terms. A FAIR workflow profile should combine specialised semantic resources, while a local vocabulary should be limited to concepts that are genuinely domain- or implementation-specific.
What is used
Consistent terms
A defined set of terms and values used consistently to describe a particular category of data, such as methods, file roles, calculation types or material properties.
Concepts and relationships
Formally represents not only terms but also relationships between materials, processes, parameters, data, software components and research results.
Persistent identifiers for units
A physical quantity is accompanied by a machine-readable
identifier for its unit, rather than only by a textual
notation such as eV or Å.
Semantic layer of a computational workflow
Different parts of a computational workflow are described using complementary specialised semantic resources.
Core semantic resources
The following resources provide a practical semantic foundation for computational materials-science workflows.
ASMO
A formal ontology for atomistic materials modelling. It covers DFT, molecular dynamics, interatomic potentials, simulation parameters and calculated physical properties.
DensityFunctionalTheoryMachineLearningPotentialMolecularDynamicsEnergyCutoffKPointMeshElasticTensor
PMD Core Ontology
A mid-level ontology for materials science and engineering. It provides common semantics for materials, processes, properties, workflows and data transformations.
workflow definition
workflow run
simulation process
workflow node
processing–structure–property relations
IUCr CIF Dictionaries
An authoritative source of standard data names for describing chemical formulae, unit-cell parameters, symmetry, space groups and atomic positions.
_chemical_formula.sum_space_group.IT_number_cell.length_a_atom_site.fract_x
W3C PROV-O
A standard model for recording which data, activities and agents participated in creating or transforming a particular workflow result.
prov:Entityprov:Activityprov:Agentprov:usedprov:wasGeneratedByprov:wasDerivedFrom
QUDT
Provides persistent machine-readable identifiers for physical quantities and units and helps avoid ambiguous free-text representations.
unit:ANGSTROMunit:EVunit:EV-PER-ANGSTROMunit:GigaPAunit:Kunit:FemtoSEC
DataCite Metadata Schema
Used at the published research-object level for titles, creators, identifiers, resource types, licences, relationships and funding information.
Identifier
Creator
Title
Subject
ResourceType
RelatedIdentifier
FundingReference
Which semantic resource to use at each workflow level
Example mapping for a modular DFT → MLIP → MD workflow.
| Description level | What needs to be described | Primary resource | Approach |
|---|---|---|---|
| Material | Chemical formula, crystal structure, cell parameters, space group and atomic sites | IUCr CIF | Reuse |
| DFT | Method, XC family, energy cutoff, k-point mesh, energy, forces and stress | ASMO | Reuse / profile |
| DFT → MLIP | DFT outputs used, quality checks and preparation of training, validation and test datasets | PROV-O + domain profile | Profile / local extension |
| MLIP | Machine-learning potential, training datasets, model configuration, model and evaluation metrics | ASMO + profile extensions | Reuse / extension |
| MLIP → MD | Selected checkpoint, deployable model and software environment | PROV-O + domain profile | Profile / extension |
| MD | Ensemble, timestep, temperature, pressure, thermostat, barostat and calculated properties | ASMO | Reuse |
| Workflow | Inputs, tasks, outputs and dependencies between computational stages | PMDco + NOMAD | Reuse / mapping |
| Provenance | Entities, activities, agents and the actual chain of result generation | W3C PROV-O | Reuse |
| Physical quantities | Length, energy, force, stress, temperature and time | QUDT | Reuse |
| Publication | DOI, creators, title, licence, related objects and funding | DataCite | Reuse / mapping |
| Package integrity | File inventory, checksums and integrity verification | SPDX | Reuse |
Recommended units for DFT → MLIP → MD
For machine interoperability, a unit should preferably be stored together with a persistent QUDT identifier.
| Physical quantity | Unit | QUDT term | Typical use |
|---|---|---|---|
| Length | Å | unit:ANGSTROM |
Lattice parameters, atomic distances, MLIP r_max |
| Energy | eV | unit:EV |
Total energy, energy errors |
| Force | eV/Å | unit:EV-PER-ANGSTROM |
DFT forces, MLIP force errors |
| Stress / modulus | GPa | unit:GigaPA |
Stress, elastic constants, elastic tensor |
| Temperature | K | unit:K |
MD temperature, applicability range |
| Time step | fs | unit:FemtoSEC |
MD timestep |
| Simulation time | ps | unit:PicoSEC |
Total MD simulation time |
{
"value": 4.5,
"unit_uri": "http://qudt.org/vocab/unit/ANGSTROM",
"unit_symbol": "Å"
}When a local extension is needed
External ontologies cover much of the core scientific semantics, but some parameters of specific software implementations or workflow transitions may require profile-specific terms.
Examples of profile-specific fields
ecutrho
r_max
interaction_layers
l_max
features
loss_weights
model_domain
split_method
deployment_method
Recommended rule
Create a local term only after checking ASMO, PMDco, MatPortal and other relevant domain semantic resources. If the concept proves useful across multiple workflows, consider aligning it with or contributing it to the appropriate international ontology community.
How to decide what to do with a metadata term
Supporting interoperability resources
NOMAD MetaInfo / workflow2
Useful for mapping computational workflows to an
inputs → tasks → outputs model and for
interoperability with materials-science data infrastructures.
SPDX
Used to formalise checksums and file-integrity information in FAIR research-object packages. SHA-256 is recommended for the MATSCI-NASU profile.
This approach supports the development of domain metadata profiles in which each field has a defined semantic source, datatype, controlled value or external IRI.
For the DFT → MLIP → MD pilot workflow, this approach has been applied in the development of a semantic crosswalk, a MATSCI-NASU controlled vocabulary and modular JSON Schemas for material, DFT, MLIP, MD, workflow, provenance and manifest metadata.