juliensimon/hypervelocity-stars-li2023

Hypervelocity Stars Catalog (Li 2023)

Credit: NASA/ESA/Hubble

Part of a dataset collection on Hugging Face.

Dataset description

The Li et al. 2023 catalog of 52 hypervelocity star (HVS) candidates — VizieR J/AJ/166/12 — combining LAMOST and Gaia data to identify stars whose total Galactocentric velocity exceeds the Milky Way escape speed under multiple plausible mass models.

Hypervelocity stars are rare tracers of extreme dynamical processes: ejection by close encounters with the central supermassive black hole (the Hills mechanism), tidal disruption of a binary in the Galactic center, dynamical ejection from a dense stellar cluster, or accretion of stars from a disrupted dwarf satellite. Confirming HVS status requires very precise 6D phase-space measurements (sky position from Gaia, proper motion from Gaia, line-of-sight velocity from spectroscopy, and distance from spectro-photometric or parallax methods) combined with an adopted Galactic potential model to compute the escape probability.

Each row records the Gaia DR3 source_id (for full-context follow-up), J2000 sky position, proper motion in both equatorial axes with uncertainties, line-of-sight velocity, heliocentric distance, the derived total Galactocentric velocity (V_GSR), and the unbound probability evaluated under five independent Milky Way potential models (Bovy 2015, McMillan 2018, Deason 2019, a baryonic-MW variant, and Williams et al. 2017) — the columns starting with 'p_unbound_' let you stress-test how a candidate's HVS classification depends on the chosen potential model. Use this dataset alongside juliensimon/gaia-dr3-spectroscopic-binaries (binary-ejection candidates), juliensimon/black-hole-catalog (for Sgr A*-context), juliensimon/cns5-nearby-stars (the local-stellar comparison sample), and juliensimon/galah-dr4-stellar-abundances (abundance-based progenitor diagnostics).

This dataset is suitable for tabular classification tasks.

Schema

Column	Type	Description	Sample	Null %
name	str	Star designation in the Li et al. 2023 catalog (e.g. 'LG-HVS1' to 'LG-HVS52'); LG = LAMOST + Gaia compilation	LG-HVS1	0.0%
gaia_source_id	int64	Gaia DR3 source_id (64-bit integer); links to full Gaia astrometry, photometry, and spectroscopy	2556679991337823488	0.0%
ra	float64	Right ascension (degrees, J2000) from Gaia DR3	12.362361	0.0%
dec	float64	Declination (degrees, J2000) from Gaia DR3	7.128955	0.0%
proper_motion_ra	float64	Proper motion in right ascension (mas/yr) including cos(Dec) factor; from Gaia DR3	11.129	0.0%
proper_motion_ra_uncertainty	float64	1-sigma uncertainty on the RA proper motion (mas/yr)	0.087	0.0%
proper_motion_dec	float64	Proper motion in declination (mas/yr) from Gaia DR3	-59.488	0.0%
proper_motion_dec_uncertainty	float64	1-sigma uncertainty on the declination proper motion (mas/yr)	0.078	0.0%
line_of_sight_velocity_kms	float64	Heliocentric radial (line-of-sight) velocity (km/s) from spectroscopy; combined with proper motion and distance to construct full 3D velocity	-345.37	0.0%
line_of_sight_velocity_uncertainty_kms	float64	1-sigma uncertainty on the line-of-sight velocity (km/s)	9.48	0.0%
heliocentric_distance_kpc	float64	Heliocentric distance (kpc); typically from spectro-photometric methods or Gaia parallax for nearer objects	2.328	0.0%
heliocentric_distance_uncertainty_kpc	float64	1-sigma distance uncertainty (kpc); dominates the velocity-budget error for most candidates	0.467	0.0%
galactocentric_velocity_kms	float64	Total velocity in the Galactic standard of rest frame (km/s); HVS candidates exceed ~530 km/s (the local escape velocity)	535.012	0.0%
galactocentric_velocity_uncertainty_kms	float64	1-sigma uncertainty on the Galactocentric velocity (km/s); a tight measurement (<10%) is essential to confirm HVS status	117.951	0.0%
p_unbound_mw_potential	float64	Posterior probability of being gravitationally unbound from the Milky Way using the standard MW potential model (Bovy 2015); >0.5 = likely escaping	0.46	0.0%
p_unbound_mcmillan2018	float64	Unbound probability using the McMillan 2018 Milky Way potential	0.45	0.0%
p_unbound_deason2019	float64	Unbound probability using the Deason 2019 Milky Way potential (higher mass)	0.58	0.0%
p_unbound_bmw_potential	float64	Unbound probability using a baryonic-MW potential variant	0.65	0.0%
p_unbound_williams2017	float64	Unbound probability using the Williams et al. 2017 Milky Way potential	0.59	0.0%

Quick stats

• 52 hypervelocity star (HVS) candidates from the Li et al. 2023 LAMOST+Gaia search
• 19 candidates with > 0.5 probability of being gravitationally unbound under the standard Milky Way potential
• Galactocentric velocities span 485 to 716 km/s (local escape speed ~530 km/s)
• Heliocentric distances: 1.2 to 4.5 kpc — most HVS lie in the outer Galactic halo
• Each candidate is scored under 5 independent Galactic potential models to assess robustness of the HVS classification

Usage

from datasets import load_dataset

ds = load_dataset("juliensimon/hypervelocity-stars-li2023", split="train")
df = ds.to_pandas()

from datasets import load_dataset
import matplotlib.pyplot as plt

df = load_dataset("juliensimon/hypervelocity-stars-li2023", split="train").to_pandas()

# Compare unbound probabilities across the 5 Galactic potential models
pot_cols = [c for c in df.columns if c.startswith("p_unbound_")]
fig, ax = plt.subplots(figsize=(8, 6))
df[pot_cols].plot(kind="box", ax=ax)
ax.axhline(0.5, color="red", linestyle="--", label="50% unbound threshold")
ax.set_ylabel("P(unbound)")
ax.set_title("HVS unbound probability under 5 MW potential models")
ax.legend()
plt.tight_layout()
plt.show()

# Top-confidence escapers under all 5 models (most robust HVS candidates)
robust = df[df[pot_cols].min(axis=1) > 0.5]
print(f"{len(robust)} stars unbound under all 5 potential models")
print(robust[["name", "galactocentric_velocity_kms", "heliocentric_distance_kpc"]])

Data source

https://cdsarc.cds.unistra.fr/viz-bin/cat/J/AJ/166/12

Related datasets

• juliensimon/gaia-dr3-spectroscopic-binaries

• juliensimon/black-hole-catalog

• juliensimon/cns5-nearby-stars

• juliensimon/galah-dr4-stellar-abundances

• juliensimon/hot-subdwarf-stars

• juliensimon/symbiotic-stars-catalog

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{hypervelocity_stars_li2023,
  title = {Hypervelocity Stars Catalog (Li 2023)},
  author = {juliensimon},
  year = {2026},
  url = {https://huggingface.co/datasets/juliensimon/hypervelocity-stars-li2023},
  publisher = {Hugging Face}
}

License

VizieR Scientific-Use Terms