Dataset Card — Scandium-Dataset v0.3.0

Dataset Description

• Name: Scandium-Dataset
• Version: v0.3.0
• Description: Curated multi-source DFT + experimental dataset for thermodynamic screening and benchmarking of battery-relevant inorganic materials, aggregated from Materials Project, OQMD, and JARVIS-DFT.
• Total entries: 267,230 (Gold: 96,242 | Validated: 140,382 | Raw: 30,108 | Experimental: 498)
• Battery subset: 82,925 entries | Electrolyte subset: 41,665 entries (strict Gold)
• Storage: Parquet (dataset/entries_v3.parquet) with indexed lookup — 0.18 GB instead of 1.6 GB JSON.
• Experimental data: 599 OBELiX entries integrated (Therrien et al. 2025, NRC-Mila), including 498 new experimental_gold tier entries with measured Li-ion conductivity.
• Transport proxies: BVSE migration barrier proxy (bvlain engine, validated against 7 known SSEs, 5/7 pass within literature ranges). See scripts/compute_bvse_barriers.py.

Sources

Source	Entries	License	Download Date
Materials Project	69,279	CC BY 4.0	2026-07-20
OQMD	171,780	Non-commercial + attribution	2026-07-20
JARVIS-DFT	25,673	CC0	2026-07-20

License Warning

⚠️ This dataset is NOT uniformly licensed. Each entry carries its own license.

• license: "CC-BY-4.0" → MP entries (commercial safe)
• license: "CC0-1.0" → JARVIS entries (commercial safe)
• license: "OQMD-noncommercial" → OQMD entries (non-commercial only, 64.4%)

Filter by the license field for programmatic usage. See LICENSE_BREAKDOWN.md.

Commercial-Safe Edition

A Commercial-Safe edition (commercial_safe_subset_v3.json) containing only MP + JARVIS entries (~94,952 entries) is available for commercial use. All entries are CC BY 4.0 or CC0 1.0 licensed. Any model trained for commercial deployment should be trained on this edition.

import json
# Load commercial-safe edition
with open("dataset/commercial_safe_subset_v3.json") as f:
    entries = json.load(f)
print(f"{len(entries):,} entries — all commercial-safe")

Fields

Primary fields

Field	Type	Description	Coverage
license	str	Per-entry license tag	100%
source_id	str	Original source ID	100%
formula	str	Raw formula string	100%
structured_formula	str	Reduced formula	100%
elements	list[str]	Element symbols	100%
nsites	int	Number of atoms in cell	100%
space_group	int	International number	99.99%
space_group_symbol	str	Hermann–Mauguin symbol	99.99%
volume	float	Cell volume (ų)	100%
density	float	Density (g/cm³)	100%
formation_energy_per_atom	float	FE (eV/atom)	100%
energy_above_hull	float	EaH (eV/atom)	90.4% → 100% (JARVIS hull added)
band_gap	float	Band gap (eV)	99.94%
total_magnetization	float	Total	100%
magnetic_ordering	str	Magnetic ordering	100%
is_metal	bool	Metallic character	100%
is_experimental	bool	From experiment?	100%
families	list[str]	Material families	100%
carrier_elements	list[str]	Mobile ion carriers	100%
anions	list[str]	Anion species	100%
structure_json	str	JSON-encoded pymatgen Structure	100%
quality_score	int	Composite quality (0–88)	100%
quality_sub_scores	dict	Per-category scores	100%
quality_flags	list[str]	Quality flags	100%
tier	str	Gold / Validated / Raw	100%
tier_detail	dict	Gate pass/fail details	100%
tier_gates	dict	Per-gate results	100%
strict_gold	dict	Strict Gold pass/fail	100%
duplicate_group	int	Dedup group ID	7.9%
provenance	dict	Full provenance chain	100%
source_weight	float	Source priority weight	100%
references	list[str]	Source references	100%

SSE proxy fields (added in v0.1.0)

Field	Type	Description	Coverage
carrier_fraction	float	Fraction of mobile ion carriers	100%
volume_per_carrier	float	Volume per mobile carrier (ų)	100%
fe_per_carrier	float	Formation energy per carrier (eV)	100%
electronic_insulation	bool	Band gap > 1.0 eV	100%
sse_family	str	SSE family classification	100%
mobile_ion	str	Primary mobile ion (Li/Na/Mg)	61.1%
oxidation_states	dict	Per-element oxidation states	100%
predicted_oxidation_states_valid	bool	BVA-validated prediction?	100%

ssb_screening block (v0.2.0)

Field	Type	Description	Coverage
mobile_ion	str	Mobile ion species	61.1%
mobile_ion_fraction	float	Fraction of mobile ions	100%
sse_family	str	SSE family tag	100%
electronic_insulation	bool	Band gap > 1.0 eV	100%
thermo_stable	bool	E_hull < 0.025 eV/atom	100%
gates_passed	list[str]	SSE screening gates passed	100%
sse_candidate_score	int	Composite SSE score (0–100)	100%
cavd_channel_dimensionality	str	0D/1D/2D/3D percolation	~61%
stability_window_low_V	float	Lower stability limit vs Li/Na	~2.2%
stability_window_high_V	float	Upper stability limit vs Li/Na	~2.2%
window_width_V	float	Electrochemical window width	~2.2%
passivating_interphase	bool	Forms passivating interphase?	~14.9%
interfacial_reaction_energy_vs_Li_eV_atom	float	Decomposition energy vs Li	~14.9%
bulk_modulus_GPa	float	Bulk modulus (geometric proxy)	100%
shear_modulus_GPa	float	Shear modulus (geometric proxy)	100%
dendrite_suppression_flag	bool	Shear modulus > 6 GPa	100%
elastic_source	str	"MP_API" / "geometric_proxy" / null	100%

Intended Use

• Primary: Training machine learning models for property prediction of solid-state battery materials
• Secondary: Materials discovery, phase stability analysis, and computational screening
• Not recommended for: Quantitative phase diagram construction (use MP/OQMD directly)

Maintenance

• Version: v0.1.0-rc.1
• DOI: pending (Zenodo archival in progress)
• Issue tracking: GitHub Issues
• Contact: Scandium Labs