Direct comparison
NOMAD vs Materials Data Facility
Compare NOMAD and Materials Data Facility: governance, hosted data types, scale, and which materials-science repository fits your project.
Side-by-side comparison
| Dimension | NOMAD | Materials Data Facility |
|---|---|---|
| What it defines | A repository and processing platform for computational materials-science data (simulation input/output) | A general publication and discovery service for structured materials datasets, computational or experimental |
| Governance | FAIR-DI e.V. (German non-profit association); development led by FAIRmat, an NFDI/DFG-funded consortium | Globus Labs (University of Chicago Dept. of Computer Science + Argonne National Laboratory), via the Computation Institute |
| Hosting | Max Planck Computing and Data Facility (MPCDF), Garching, Germany | Multi-petabyte storage at Argonne National Laboratory and NCSA (UIUC), transfer via Globus |
| Origin | 2013-2014, Humboldt-Universität zu Berlin and the Fritz Haber Institute of the Max Planck Society | Launched in support of the US Materials Genome Initiative, with CHiMaD (NIST-funded center) as a founding partner |
| Primary data type | Raw/derived computational data: DFT, GW, molecular dynamics, and other simulation output | Any structured materials dataset — computational, experimental, or mixed |
| Code-aware parsing | Yes — automatically parses and normalizes files from 60+ simulation codes into a common schema | No dedicated code parsing — relies on depositor-supplied metadata |
| Scale (indicative) | On the order of 19 million uploaded entries covering several million distinct materials (grows continuously) | Thousands of published datasets; scale reported per-dataset rather than per-entry |
| Search strength | Deep property-level search across normalized simulation output; OPTIMADE API for cross-database structure queries | Dataset-level discovery driven by depositor metadata |
| Large-file / HPC transfer | Standard web upload/API | Globus-based, built for high-throughput, resumable transfer of large, high-file-count datasets |
| Typical best fit | Depositing/searching raw simulation output; training ML models on normalized data | Publishing a citable, DOI'd, mixed computational+experimental dataset; moving large datasets to/from HPC |
| Relationship to MGI | Independent European effort, frequently cited alongside MGI databases as a parallel FAIR-for-materials initiative | Built directly with Materials Genome Initiative funding and CHiMaD partnership |
Common questions
FAQ
Can I use both NOMAD and MDF for the same project?+
Yes, and it's common practice. Many researchers deposit raw simulation output in NOMAD for its automatic parsing and cross-code search, then publish a curated, DOI'd dataset bundle (which may reference the NOMAD entries) through MDF for citation in a paper's data-availability statement.
Which one should I list in a Data Management Plan for computational materials-science work?+
If your output is primarily raw simulation files from a supported code, list NOMAD as the primary repository. If you need a single DOI for a mixed or experimental dataset, or you're already using Globus for transfer, list MDF. Both are indexed in re3data and are recognized domain repositories for materials science.
Are NOMAD and MDF free to use?+
Yes, both are free for researchers to deposit and publish data. NOMAD is funded through FAIR-DI e.V./FAIRmat and German research-infrastructure funding; MDF runs on Globus's non-profit infrastructure, which offers free core services for academic use.
Going deeper
