科研 Skill

材料数据库跨库查询

materials-databases

Covers querying public materials databases beyond Materials Project: AFLOW, OQMD, NOMAD, JARVIS. Use when pulling or comparing data from these live, API-integrated resources.

发布者 Codar安全审核通过v0.1.0更新于 2026-11-18

SKILL.md

Materials Databases

Overview

Unlike neuroscience's mostly-passive public datasets, several materials databases are live APIs that a workflow queries directly and repeatedly, not a corpus downloaded once — hence this being a full skill rather than an appendix to a data-standards skill. Materials Project queries are covered in materials-structure-analysis/references/materials-project-queries.md; this skill covers the other major databases and how to compare results across them.

When to use this skill

Activate when the request involves:

  • AFLOW, OQMD, Open Quantum Materials Database, NOMAD, JARVIS, ICSD, Matbench
  • Terms: high-throughput database, DFT-computed properties (from a database, not your own calculation), benchmark dataset, FAIR data
  • "Query AFLOW/OQMD/NOMAD/JARVIS for...," "compare this property across databases," "find a benchmark dataset"

Core usage

AFLOW

from aflow import search, K

results = search(catalog="icsd").filter(K.species == ("Fe", "O")).select(K.energy_atom, K.spacegroup_relax)
for entry in results:
    print(entry.auid, entry.energy_atom, entry.spacegroup_relax)

OQMD (Open Quantum Materials Database)

import requests

response = requests.get(
    "http://oqmd.org/oqmdapi/formationenergy",
    params={"composition": "Fe2O3", "fields": "name,delta_e,spacegroup"},
)
entries = response.json()["data"]

NOMAD

import requests

response = requests.post(
    "https://nomad-lab.eu/prod/v1/api/v1/entries/query",
    json={"query": {"results.material.elements": {"all": ["Fe", "O"]}}, "pagination": {"page_size": 20}},
)
entries = response.json()["data"]

JARVIS (NIST)

from jarvis.db.figshare import data as jarvis_data

dft_3d = jarvis_data("dft_3d")   # downloads/caches the JARVIS-DFT 3D materials dataset
fe_containing = [entry for entry in dft_3d if "Fe" in entry["formula"]]

Matbench — standardized ML benchmark suite

from matbench.bench import MatbenchBenchmark

mb = MatbenchBenchmark(autoload=False)
for task in mb.tasks:
    task.load()   # loads a specific standardized materials-property-prediction benchmark

Validation & Pitfalls

Canonical references: Curtarolo et al. (2012), "AFLOW: An automatic framework for high-throughput materials discovery," Computational Materials Science; Saal et al. (2013), "Materials design and discovery with high-throughput density functional theory: the Open Quantum Materials Database (OQMD)," JOM; Draxl & Scheffler (2019), "The NOMAD laboratory: from data sharing to artificial intelligence," Journal of Physics: Materials; Choudhary et al. (2020), "The joint automated repository for various integrated simulations (JARVIS) for data-driven materials design," npj Computational Materials.

  • Different databases use different DFT methodologies (functional, pseudopotentials, convergence settings), so a property value for the "same" material can differ meaningfully across databases — this is not noise or an error in one of them, it's a real methodology difference. Never mix entries from different databases into the same phase-diagram or stability analysis (same methodology-consistency principle as materials-structure-analysis/references/phase-diagrams.md); cross-database comparison is useful for sanity-checking a value's robustness, not for building a single consistent dataset.
  • API schemas and endpoints for these databases change over time, sometimes without long deprecation notices — code that worked against a database's API a year ago may fail or return differently structured data today; verify against current documentation rather than assuming a cached example (including the ones above) is still exactly correct.
  • Database coverage is not uniform across chemical space — some databases emphasize inorganic crystals (Materials Project, OQMD, AFLOW), others are broader in scope (NOMAD accepts many calculation types and codes) or benchmark-focused (Matbench). Check whether a database actually covers the chemistry/property of interest before assuming an absence of results means the material is unstudied, rather than just absent from that particular database.
  • A high-throughput database entry represents one calculation, not a validated, peer-reviewed result — treat database-sourced values with the same skepticism as any single calculation (see computational-materials/references/dft.md's convergence/methodology caveats), not as ground truth simply because it's in a large curated database.
  • Rate limits and bulk-query etiquette matter for these APIs — an unbounded or high-frequency query pattern can get an API key throttled or blocked; bound queries explicitly and check each database's documented rate limits before running a large automated query campaign.

示例提示词

我想跨AFLOW、OQMD和NOMAD查询并比较同一化合物的计算结果。

所属插件

材料表征、性能与材料信息学面向实验与数据驱动的材料研究:XRD/电镜/谱学/热分析/力学测试等表征,合成与加工,电化学与半导体表征,相场、CALPHAD 与连续介质力学,材料数据库、特征工程、贝叶斯优化与主动学习,选材与生命周期评价。

安装该插件会一并获得同领域的 42 个 Skill / MCP。

同插件的其他科研 Skill

常见问题

材料数据库跨库查询能做什么?

Covers querying public materials databases beyond Materials Project: AFLOW, OQMD, NOMAD, JARVIS. Use when pulling or comparing data from these live, API-integrated resources。

如何安装材料数据库跨库查询?

在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「材料数据库跨库查询」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。

材料数据库跨库查询需要哪些权限?

该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。

材料数据库跨库查询经过官方审核吗?

是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。