juliensimon/gaia-dr3-chemical-cartography

---
license: other
license_name: cc-by-nc-3.0-igo
license_link: https://creativecommons.org/licenses/by-nc/3.0/igo/
pretty_name: "Gaia DR3 Chemical Cartography"
language:
- en
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 Rad"
task_categories:
- tabular-classification
- tabular-regression
tags:
- space
- gaia
- milky-way
- galactic-dynamics
- stellar-kinematics
- esa
- astronomy
- open-data
- tabular-data
- parquet
size_categories:
- 1M<n<10M
configs:
- config_name: default
data_files:
- split: train
path: data/gaia_dr3_chemical_cartography.parquet
default: true
---

# Gaia DR3 Chemical Cartography

<div align="center">
<img src="banner.jpg" alt="The Milky Way galaxy seen from above — NASA composite" width="400">
<p><em>Credit: NASA/JPL-Caltech</em></p>
</div>

*Part of a [dataset collection](https://huggingface.co/collections/juliensimon/astronomy-datasets-69c24caf2f17e36128946743) 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

```python
from datasets import load_dataset

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

```python
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](https://huggingface.co/datasets/juliensimon/gaia-dr3-young-stellar-objects)

- [juliensimon/galah-dr4-stellar-abundances](https://huggingface.co/datasets/juliensimon/galah-dr4-stellar-abundances)

- [juliensimon/apogee-dr17](https://huggingface.co/datasets/juliensimon/apogee-dr17)

- [juliensimon/rave-dr6](https://huggingface.co/datasets/juliensimon/rave-dr6)

> If you find this dataset useful, please consider [giving it a like](https://huggingface.co/datasets/juliensimon/gaia-dr3-chemical-cartography) on Hugging Face. It helps others discover it.

## About the author

Created by [Julien Simon](https://julien.org) — AI Operating Partner at Fortino Capital. Part of the [Space Datasets](https://julien.org/datasets) collection.

## Citation

```bibtex
@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](https://creativecommons.org/licenses/by-nc/3.0/igo/)