Gaia DR3 high radial velocity stars: Genuine fast-moving objects or outliers?

This study validates the radial velocities of 134 high-velocity stars from Gaia DR3 using ground-based spectroscopy, confirming that most are genuine accreted objects on retrograde orbits associated with past merging events while identifying a contamination rate of spurious measurements in the low signal-to-noise regime.

The Gist

Imagine the Gaia mission as a massive, high-tech cosmic census taker. Its job is to map the Milky Way, measuring the position, brightness, and speed of over a billion stars. In its third major data dump (Gaia DR3), it released a list of 33.8 million stars with measured speeds.

However, there was a catch. To get to the faintest, dimmest stars, Gaia had to look at very "noisy" data. It's like trying to hear a whisper in a hurricane. Sometimes, the static (noise) looks like a voice, and the computer gets tricked into thinking a star is moving incredibly fast when it's actually just sitting there or moving normally.

This paper is the story of a team of astronomers acting as cosmic detectives to figure out which of these "fast-moving" stars are real and which are just optical illusions.

The Mystery: The "Fast" Stars

The astronomers were interested in High Radial Velocity (HRV) stars. These are stars moving so fast (over 500 km/s, or about 1.1 million mph!) that they seem to be zooming away from or toward us at breakneck speeds.

In the Gaia list, there were about 1,500 of these "super-speeders." But the team suspected many were fakes—ghosts created by bad data.

The Investigation: Ground Truth

To solve the mystery, the team didn't just trust the space telescope. They grabbed their own "flashlights" (massive ground-based telescopes in France and Chile) and took a closer look at 134 of these suspicious stars.

Think of Gaia's data as a blurry, low-resolution photo taken from a distance. The ground-based telescopes are like high-definition zoom lenses.

The Results:

• The Real Deal: For 104 of the stars, the ground telescopes confirmed Gaia was right. These stars are indeed zooming through space at incredible speeds.
• The Fakes: For 30 stars, the ground telescopes said, "Nope, you're not moving that fast." Gaia had been fooled by the noise. These stars were actually moving at normal speeds, but the data looked like a glitch.

The "Signal-to-Noise" Rule

The team discovered a simple rule of thumb to tell the fakes from the real ones: The Signal-to-Noise Ratio (S/N).

• High S/N (Clear Signal): If the data is clear and bright, Gaia is almost always right.
• Low S/N (Noisy Signal): If the data is faint and grainy (S/N below 7), the chance of a fake "fast star" skyrockets.
  • Analogy: Imagine trying to read a sign in a foggy storm. If the sign is bright and close, you can read it perfectly. If it's dim and far away, your brain might fill in the gaps and think you see a word that isn't there. In the "foggiest" data (S/N between 2 and 3), 83% of the "fast stars" were actually fakes!

The Twist: The "Retrograde" Rebels

Once the team filtered out the fakes and kept only the confirmed fast stars, they looked at where these stars were going.

Most stars in our galaxy, including our Sun, are like cars in a roundabout, all spinning in the same direction (counter-clockwise). But the team found that most of these genuine fast stars are driving the wrong way.

• The Analogy: Imagine a highway where everyone is driving north. Suddenly, you see a fleet of cars speeding south at 200 mph. That's what these stars are doing. They are on retrograde orbits (moving opposite to the galaxy's spin).

The Origin Story: Cosmic Immigrants

Why are they driving the wrong way? The team looked at their energy and paths and realized something exciting: These stars are cosmic immigrants.

They weren't born in the Milky Way. They were likely stolen from smaller, dwarf galaxies that crashed into our galaxy billions of years ago. When these smaller galaxies merged with the Milky Way, their stars were flung into strange, high-speed, backward orbits.

• The "Fossils": These stars are like living fossils. By studying them, we are essentially reading the history books of how our galaxy was built, piece by piece, from smaller collisions.

The Conclusion

1. Be Careful: If you see a star in the Gaia data moving super fast but the data is very "noisy" (low S/N), it's probably a glitch.
2. The Real Fast Stars: The ones that passed the test are real, and they are mostly "rebels" moving backward.
3. Galactic History: These rebels are the survivors of ancient galactic crashes, giving us a direct look at how the Milky Way grew up.

In short, the astronomers cleaned up the "noise" in the cosmic census, found the real speedsters, and discovered that our galaxy is a patchwork quilt stitched together from the wreckage of smaller galaxies.

Technical Summary

1. Problem Statement

The third Gaia data release (DR3) contains 33.8 million radial velocity (RV) measurements extending to a magnitude of $G_{RVS} = 14$. To achieve this depth, spectra were processed down to very low signal-to-noise ratios (S/N), as low as 2. In this low S/N regime, noise-induced secondary peaks in the cross-correlation function (CCF) can exceed the true primary peak, leading to spurious radial velocity determinations.

While quality filters were applied to the full dataset, the population of High Radial Velocity (HRV) stars (defined here as $|RV| > 500$ km s$^{-1}$) is extremely sparse. Consequently, even a small number of erroneous measurements (a few hundred) can significantly contaminate this specific subset, potentially mimicking high-velocity stars (HVSs) or hyper-velocity stars that do not exist. The study aims to distinguish genuine fast-moving objects from these outliers.

2. Methodology

The authors employed a multi-step approach combining ground-based spectroscopy with existing public surveys:

• Target Selection: From the 1,544 Gaia DR3 stars with $|RV| > 500$ km s$^{-1}$, a sample of 134 targets was selected. The selection ensured a balanced distribution of positive and negative velocities and spanned a wide range of Gaia S/N (from 2 to >100).
• Ground-Based Observations:
  • SOPHIE (Observatoire de Haute-Provence, France): 72 stars observed using a high-efficiency mode (R $\approx$ 39,000).
  • UVES (VLT, Chile): 70 stars observed (8 stars were observed by both instruments) using high-resolution settings (R $\approx$ 90,000–110,000).
  • Analysis: RVs were derived using two methods:
    1. Cross-correlation: Using SOPHIE masks matched to stellar temperatures.
    2. Template-matching: Using synthetic spectra (SYNTHE/ATLAS9) to handle broad lines (e.g., Balmer lines) where cross-correlation might fail.
• Cross-Matching with Public Surveys: The sample was cross-matched with five major ground-based spectroscopic surveys (APOGEE DR17, GALAH DR3, GES DR3, LAMOST DR7, RAVE DR6) to create a larger validation set of ~2.7 million stars.
• Orbital Integration: A "clean" sample of HRV stars (confirmed ground-based RVs + high S/N Gaia DR3 stars) was integrated using an axisymmetric Galactic gravitational potential (Pouliaisis et al. 2017) to determine orbital parameters (eccentricity, energy, angular momentum).
• Age Determination: Ages were derived for a subset of 74 stars using the SPInS code and BaSTI evolutionary tracks.

3. Key Contributions

• Validation of Gaia DR3 HRVs: The first systematic ground-based verification of ~100 high-velocity candidates from Gaia DR3.
• Quantification of Outlier Rates: Establishment of a statistical relationship between the rate of spurious RV measurements, S/N, and the absolute value of the radial velocity.
• Kinematic Characterization: Detailed orbital analysis of a clean sample of HRV stars, linking them to specific accretion events in the Milky Way's history.
• Identification of Contamination Sources: Demonstration that low S/N is the primary driver of errors in the HRV tail, rather than contamination by nearby bright stars.

4. Results

A. Confirmation and Refutation of RVs

• Confirmation: Ground-based measurements confirmed the Gaia DR3 RVs for 104 out of 134 targets (within a $\pm 50$ km s$^{-1}$ tolerance).
• Refutation: The remaining 30 targets were refuted. All 30 refuted stars had Gaia DR3 S/N below 7.
• Nature of Errors: The refuted stars typically showed ground-based velocities near 0 km s$^{-1}$ (or typical halo values), while Gaia reported extreme values ($>500$ km s$^{-1}$). In some cases, the stars were identified as binaries or variables with emission lines that confused the Gaia pipeline.

B. Outlier Rate Analysis

By combining the new data with public surveys, the authors quantified the "outlier rate" (spurious measurements) as a function of S/N and velocity interval:

• S/N Dependence: The error rate is strongly correlated with S/N.
  • For S/N in the range [2, 3) and velocity interval [400, 1000) km s$^{-1}$, the outlier rate reaches 83%.
  • For S/N > 7, the outlier rate drops significantly, and all targets in the study were confirmed.
• Velocity Dependence: The contamination is highest in the "wings" of the velocity distribution (high absolute velocities) because the true population is sparse, making the flat distribution of noise peaks more dominant.
• Spatial Distribution: Spurious stars are not disproportionately clustered in the Galactic plane, suggesting that neighbor contamination is not the primary cause; low S/N is the dominant factor.

C. Kinematic and Dynamic Properties

The clean sample (108 confirmed + 53 high S/N unconfirmed) reveals:

• Retrograde Orbits: The vast majority (92.5%) of HRV stars have retrograde azimuthal velocities ($V_{\phi} < -100$ km s$^{-1}$). This is a selection effect: stars on retrograde orbits move opposite to the Sun, making their projected radial velocities appear larger.
• Orbital Parameters:
  • Median vertical angular momentum ($L_Z$): $-2.2 \times 10^3$ kpc km s$^{-1}$.
  • Most stars are gravitationally bound (negative energy) but on highly eccentric orbits ($e > 0.4$).
  • Pericenters are mostly > 1 kpc, though 6 stars have pericenters < 1 kpc.
• Accretion History: In the Energy vs. $L_Z$ diagram, the stars cluster in regions associated with known accretion events: Sequoia, Arjuna, I'itoi, Antaeus, ED-2, and ED-3. This strongly suggests these HRV stars are remnants of past mergers.
• Potential Hypervelocity Stars (HVSs): Four stars in the sample showed positive total energy (unbound). However, these were all from the unconfirmed group (S/N > 10 but no ground truth). Further analysis suggested their velocities might be biased by template mismatches or poor astrometric fits, requiring future confirmation.

D. Age Properties

• Bimodal Age Distribution:
  • Old Population (>7–8 Ga): Consistent with the accreted halo population.
  • Young Population (<6 Ga): Likely Blue Stragglers (BSS) or evolved BSS formed in binary systems, or potentially young metal-poor stars formed in interacting dwarf galaxies.

5. Significance

• Data Quality Warning: The study provides a critical warning for users of Gaia DR3. High radial velocity values ($|RV| > 500$ km s$^{-1}$) derived from spectra with S/N < 7 are highly unreliable (up to 83% error rate in the highest velocity bin).
• Scientific Utility: By filtering out these outliers, the authors provide a robust sample of genuine high-velocity stars. This sample confirms that the "fast-moving" tail of the Gaia RV distribution is dominated by accreted retrograde stars rather than stars ejected from the Galactic center or the Large Magellanic Cloud.
• Future Outlook: The findings highlight the challenges of extending RV measurements to fainter magnitudes (Gaia DR4/DR5). The authors note that future releases will need advanced machine learning techniques to better distinguish noise peaks from true signals in low S/N regimes.

In conclusion, this paper successfully separates genuine astrophysical fast-moving objects from instrumental artifacts in Gaia DR3, refining our understanding of the Milky Way's accretion history and providing essential quality control metrics for future high-velocity star searches.