mat-db-optimade

Goal

To query experimental crystal structures (e.g. from the Crystallography Open Database - COD, NOMAD) using the standardized REST API provided by OPTIMADE, and to extract the structural data and metadata in a JSON format.

Instructions

Step 1. Querying an OPTIMADE Database

Use the query_optimade.py script to fetch structure data based on standard OPTIMADE query language filters.

# Env: base-agent
python .agents/skills/mat-db-optimade/scripts/query_optimade.py \
    results.json \
    --filter 'elements HAS ALL "Na", "Cl"' \
    --provider cod \
    --max_results 5

Parameters:

• output: Path to the output JSON file where results will be stored.
• --filter: The OPTIMADE filter string. (e.g., elements HAS ALL "O", "Ti").
• --provider: Comma-separated list of database providers. Defaults to cod.
• --max_results: Maximum number of structures to pull per provider. Defaults to 10.

Supported Database Providers (Common short IDs)

When using the --provider parameter, you can provide the following short string identifiers for well-known databases:

• cod – Crystallography Open Database: The largest open-access collection of experimentally determined crystal structures.
• nmd – NOMAD: Comprehensive repository of both theoretical methods and experimental configurations.
• aflow – AFLOW: Automatic-FLOW Computational Materials Data Repository.
• mp – Materials Project: Well-known theoretical stable/unstable materials repository based on DFT.
• oqmd – Open Quantum Materials Database: High-throughput DFT database of thermodynamic and structural properties.
• jarvis – JARVIS: Joint Automated Repository for Various Integrated Simulations.

Constraints

• Environments: The scripts require the base-agent Conda environment.
• Provider Status: Sometimes specific providers (like Aflow or NOMAD) may experience API downtime. If a provider fails, the optimade client will log warnings.

References

• Andersen, C. W., et al. "OPTIMADE, an API for exchanging materials data." Scientific Data, 8, 217 (2021). DOI

---

Author: Bowen Deng Contact: GitHub @learningmatter-mit