juliensimon/gaia-dr3-chemical-cartography

Gaia DR3 Chemical Cartography

Credit: NASA/JPL-Caltech

Part of a dataset collection on Hugging Face.

Dataset description

The Gaia DR3 Chemical Cartography catalog provides galactic orbital parameters for approximately 5.6 million stars derived from Gaia astrometry, radial velocities, and chemical abundances from the Radial Velocity Spectrometer (RVS). Each star's orbit in the Milky Way's gravitational potential is fully characterized by action integrals (Jr, Jz, Jphi), orbital extremes (rapo, rperi, zmax), eccentricity, and current phase-space position in both cylindrical and Cartesian Galactocentric coordinates. All parameters are provided with three confidence bounds (median, upper, lower 1-sigma) that propagate uncertainties from proper motion, parallax, and radial velocity measurements through the orbital integration.

Actions are the most powerful coordinates for galactic archaeology because they are conserved (or nearly so) over many orbital periods in a smooth potential. Different stellar populations occupy distinct, non-overlapping regions of action space: the thin disk concentrates near (Jr ≈ 0, Jz ≈ 0, |Jphi| ≈ 1500–2000 kpc·km/s), the thick disk spreads to higher Jr and Jz at similar Jphi, accreted halo stars from disrupted dwarf galaxies form kinematic streams at characteristic (Jr, Jz, Jphi) loci, and the in-situ halo occupies high-eccentricity (ecc > 0.7) retrograde orbits. Combining orbital actions with chemical abundances ([Fe/H], [α/Fe]) from the same spectra enables chemo-dynamical dissection of the Galaxy's assembly history — the primary science goal of chemical cartography.

This is the largest kinematic catalog ever assembled for Milky Way stars, representing an order-of-magnitude improvement over previous surveys such as RAVE DR6 (450,000 stars), GALAH DR4 (600,000 stars), or APOGEE DR17 (~700,000 stars). The Toomre diagram (sqrt(vz² + vR²) vs vφ) cleanly separates thin disk, thick disk, and halo populations. Scatter plots of zmax vs rapo or ecc distributions reveal the relative contributions of in-situ and accreted material across the Galaxy.

This dataset is suitable for tabular classification, tabular regression tasks.

Schema

Column	Type	Description	Sample	Null %
source_id	int64	Gaia DR3 unique source identifier; use for cross-matching with other Gaia tables	3332894779520	0.0%
jr_med	float64	Radial galactic orbital action Jr, median estimate (kpc·km/s); measures radial oscillation amplitude in the Milky Way potential	0.01154261	1.1%
jz_med	float64	Vertical galactic orbital action Jz, median estimate (kpc·km/s); measures oscillation amplitude perpendicular to the Galactic plane	0.0024616383	1.1%
jphi_med	float64	Azimuthal action Jφ (angular momentum), median estimate (kpc·km/s); conserved for axisymmetric potentials; large negative values indicate prograde disk orbits	0.93710554	1.1%
rplane_med	float64	Current Galactocentric cylindrical radius R, median (kpc); distance from the Galactic rotation axis	8.617176	1.1%
vrplane_med	float64	Galactocentric radial velocity vR in the plane, median (km/s); positive = moving away from Galactic center	-13.085062	1.1%
vz_med	float64	Galactocentric vertical velocity vz, median (km/s); positive = moving toward north Galactic pole	-2.1960492	1.1%
vphi_med	float64	Galactocentric azimuthal velocity vφ, median (km/s); negative values indicate prograde (disk-like) orbits under the Milky Way convention	213.95734	1.1%
zmax_med	float64	Maximum height above the Galactic plane reached during the orbit, median (kpc); thin disk stars have zmax < 0.3 kpc, thick disk 0.3–3 kpc, halo > 3 kpc	0.40377092	1.1%
rapo_med	float64	Orbital apocentric radius (farthest point from Galactic center), median (kpc)	8.680873	1.1%
rperi_med	float64	Orbital pericentric radius (closest approach to Galactic center), median (kpc)	6.680762	1.1%
ecc_med	float64	Orbital eccentricity, median (0=circular, 1=radial); thin disk: ecc < 0.2, thick disk: 0.2–0.5, halo: > 0.5	0.13019727	1.1%
x_med	float64	Galactocentric Cartesian x coordinate, median (kpc); Sun is at x ≈ -8.3 kpc	8.617152	1.1%
y_med	float64	Galactocentric Cartesian y coordinate, median (kpc)	-0.019991076	1.1%
z_med	float64	Galactocentric Cartesian z coordinate, median (kpc); z = 0 is the Galactic plane	-0.39784664	1.1%
energy_med	float64	Total orbital energy (gravitational + kinetic), median (km²/s²); negative values indicate gravitationally bound orbits	-166490.28	1.1%
jr_hi	float64	Radial action Jr, upper 1-sigma confidence bound (kpc·km/s)	0.011789521	1.1%
jz_hi	float64	Vertical action Jz, upper 1-sigma confidence bound (kpc·km/s)	0.0025126631	1.1%
jphi_hi	float64	Azimuthal action Jφ, upper 1-sigma confidence bound (kpc·km/s)	0.93831515	1.1%
rplane_hi	float64	Galactocentric radius R, upper 1-sigma confidence bound (kpc)	8.621864	1.1%
vrplane_hi	float64	Radial velocity vR, upper 1-sigma confidence bound (km/s)	-12.271209	1.1%
vz_hi	float64	Vertical velocity vz, upper 1-sigma confidence bound (km/s)	-1.3347614	1.1%
vphi_hi	float64	Azimuthal velocity vφ, upper 1-sigma confidence bound (km/s)	214.29321	1.1%
zmax_hi	float64	Maximum Galactic height zmax, upper 1-sigma confidence bound (kpc)	0.40878028	1.1%
rapo_hi	float64	Apocentric radius rapo, upper 1-sigma confidence bound (kpc)	8.6835165	1.1%
rperi_hi	float64	Pericentric radius rperi, upper 1-sigma confidence bound (kpc)	6.6999903	1.1%
ecc_hi	float64	Orbital eccentricity, upper 1-sigma confidence bound	0.13163313	1.1%
x_hi	float64	Cartesian x coordinate, upper 1-sigma confidence bound (kpc)	8.6218405	1.1%
y_hi	float64	Cartesian y coordinate, upper 1-sigma confidence bound (kpc)	-0.019845717	1.1%
z_hi	float64	Cartesian z coordinate, upper 1-sigma confidence bound (kpc)	-0.3948026	1.1%
jr_lo	float64	Radial action Jr, lower 1-sigma confidence bound (kpc·km/s)	0.011182615	1.1%
jz_lo	float64	Vertical action Jz, lower 1-sigma confidence bound (kpc·km/s)	0.002410012	1.1%
jphi_lo	float64	Azimuthal action Jφ, lower 1-sigma confidence bound (kpc·km/s)	0.93547726	1.1%
rplane_lo	float64	Galactocentric radius R, lower 1-sigma confidence bound (kpc)	8.614498	1.1%
vrplane_lo	float64	Radial velocity vR, lower 1-sigma confidence bound (km/s)	-13.60701	1.1%
vz_lo	float64	Vertical velocity vz, lower 1-sigma confidence bound (km/s)	-3.0250373	1.1%
vphi_lo	float64	Azimuthal velocity vφ, lower 1-sigma confidence bound (km/s)	213.46265	1.1%
zmax_lo	float64	Maximum Galactic height zmax, lower 1-sigma confidence bound (kpc)	0.3993213	1.1%
rapo_lo	float64	Apocentric radius rapo, lower 1-sigma confidence bound (kpc)	8.671579	1.1%
rperi_lo	float64	Pericentric radius rperi, lower 1-sigma confidence bound (kpc)	6.662339	1.1%
ecc_lo	float64	Orbital eccentricity, lower 1-sigma confidence bound	0.12811361	1.1%
x_lo	float64	Cartesian x coordinate, lower 1-sigma confidence bound (kpc)	8.614475	1.1%
y_lo	float64	Cartesian y coordinate, lower 1-sigma confidence bound (kpc)	-0.020245597	1.1%
z_lo	float64	Cartesian z coordinate, lower 1-sigma confidence bound (kpc)	-0.40317678	1.1%
ecc_uncertainty	float64	Average 1-sigma eccentricity uncertainty: (ecc_hi - ecc_lo) / 2; reflects propagated errors from astrometry and radial velocity	0.001759759999999999	1.1%
is_halo_candidate	bool	Boolean flag: True when ecc_med > 0.7 AND |zmax_med| > 3.0 kpc, indicating likely halo or accreted stellar population	False	0.0%

Quick stats

• 5,591,594 stars with galactic orbital parameters
• Median orbital eccentricity (ecc_med): 0.130
• Fraction with ecc > 0.5 (eccentric/halo orbits): 2.8%
• Median maximum Galactic height (zmax_med): 0.372 kpc
• Median apocentric radius (rapo_med): 8.671 kpc
• Halo candidates (ecc > 0.7 & |zmax| > 3 kpc): 25,056

Usage

from datasets import load_dataset

ds = load_dataset("juliensimon/gaia-dr3-chemical-cartography", split="train")
df = ds.to_pandas()

from datasets import load_dataset
import numpy as np

ds = load_dataset("juliensimon/gaia-dr3-chemical-cartography", split="train")
df = ds.to_pandas()

# Toomre diagram: separates thin disk, thick disk, and halo
import matplotlib.pyplot as plt
vtot = np.sqrt(df["vz_med"]**2 + df["vrplane_med"]**2)
plt.hexbin(df["vphi_med"], vtot, gridsize=300, mincnt=1, cmap="hot")
plt.colorbar(label="Count")
plt.xlabel("vφ (km/s)")
plt.ylabel("√(vz² + vR²) (km/s)")
plt.title("Toomre Diagram — Gaia DR3 Chemical Cartography")
plt.show()

# Eccentricity distribution
df["ecc_med"].hist(bins=100, log=True)
plt.xlabel("Orbital eccentricity")
plt.ylabel("Count (log)")
plt.title("Orbital Eccentricity Distribution")
plt.show()

# zmax vs rapo to show thin/thick disk/halo separation
thin_disk = df[df["ecc_med"] < 0.2]
thick_disk = df[(df["ecc_med"] >= 0.2) & (df["ecc_med"] < 0.5)]
halo = df[df["ecc_med"] >= 0.7]
plt.scatter(thin_disk["rapo_med"].sample(5000), thin_disk["zmax_med"].sample(5000),
            s=1, alpha=0.3, label="Thin disk (ecc<0.2)")
plt.scatter(thick_disk["rapo_med"].sample(5000), thick_disk["zmax_med"].sample(5000),
            s=1, alpha=0.3, label="Thick disk (0.2≤ecc<0.5)")
plt.scatter(halo["rapo_med"].sample(min(5000, len(halo))),
            halo["zmax_med"].sample(min(5000, len(halo))),
            s=1, alpha=0.5, label="Halo (ecc≥0.7)")
plt.xlabel("Apocentric radius rapo (kpc)")
plt.ylabel("Maximum height zmax (kpc)")
plt.legend()
plt.title("Galactic Structure: zmax vs rapo")
plt.show()

# Action space: Jr vs Jphi colored by Jz
halo_candidates = df[df["is_halo_candidate"]]
print(f"Halo candidates: {len(halo_candidates):,}")

Data source

https://gea.esac.esa.int/archive/

Related datasets

• juliensimon/gaia-dr3-young-stellar-objects

• juliensimon/galah-dr4-stellar-abundances

• juliensimon/apogee-dr17

• juliensimon/rave-dr6

If you find this dataset useful, please consider giving it a like on Hugging Face. It helps others discover it.

About the author

Created by Julien Simon — AI Operating Partner at Fortino Capital. Part of the Space Datasets collection.

Citation

@dataset{gaia_dr3_chemical_cartography,
  title = {Gaia DR3 Chemical Cartography},
  author = {juliensimon},
  year = {2026},
  url = {https://huggingface.co/datasets/juliensimon/gaia-dr3-chemical-cartography},
  publisher = {Hugging Face}
}

License

CC-BY-NC-3.0-IGO