[Paper Review] Gaia Early Data Release 3: Parallax bias versus magnitude, colour, and position

L. Lindegren, U. Bastian+15 more|arXiv (Cornell University)|Dec 3, 2020

Stellar, planetary, and galactic studies32 references41 citations

TL;DR

The paper maps Gaia EDR3 parallax bias (zero point) as a function of magnitude, colour, and position for both five- and six-parameter solutions, using quasars, LMC stars, and other differential methods to guide a multi-parameter correction model.

ABSTRACT

Gaia Early Data Release 3 (Gaia EDR3) gives trigonometric parallaxes for nearly 1.5 billion sources. Inspection of the EDR3 data for sources identified as quasars reveals that their parallaxes are biased, that is systematically offset from the expected distribution around zero, by a few tens of microarcsec. We attempt to map the main dependencies of the parallax bias in EDR3. In principle this could provide a recipe for correcting the EDR3 parallaxes. For faint sources the quasars provide the most direct way to estimate parallax bias. In order to extend this to brighter sources and a broader range of colours, we use differential methods based on physical pairs (binaries) and sources in the Large Magellanic Cloud. The functional forms of the dependencies are explored by mapping the systematic differences between EDR3 and DR2 parallaxes. The parallax bias is found to depend in a non-trivial way on (at least) the magnitude, colour, and ecliptic latitude of the source. Different dependencies apply to the five- and six-parameter solutions in EDR3. While it is not possible to derive a definitive recipe for the parallax correction, we give tentative expressions to be used at the researcher's discretion and point out some possible paths towards future improvements.

Motivation & Objective

• Quantify the main dependencies of the Gaia EDR3 parallax bias (zero point) on magnitude, colour, and position.
• Develop functional forms Z5 and Z6 for five- and six-parameter solutions to describe the bias.
• Provide practical, tentative correction expressions for researchers to apply in astrophysical analyses.

Proposed method

• Use quasars as direct probes of parallax bias at faint magnitudes and broader colour ranges.
• Extend bias mapping to brighter stars via differential methods using physical pairs (binaries) and Large Magellanic Cloud data.
• Model the bias as a multi-parameter function Z(G, nu_eff, beta) with a basis-function expansion across magnitude, colour, and ecliptic latitude.
• Fit 195 parameters (13 x 5 x 3) with an iterative overfitting control to obtain a tractable model.
• Compare EDR3 and DR2 parallaxes to inform the structure of the bias and to guide the functional form of Z5 and Z6.
• Present a practical, albeit tentative, correction framework and discuss limitations and avenues for improvement.

Experimental results

Research questions

• RQ1 What are the primary dependencies of Gaia EDR3 parallax bias on magnitude (G), effective wavenumber (nu_eff), and ecliptic latitude (beta) for five- and six-parameter solutions?
• RQ2 Can a multi-parameter correction function Z5 and Z6 be constructed to approximate the parallax zero point across the EDR3 catalogue, and how well does it perform when validated against external indicators (quasars, LMC)?
• RQ3 How does the EDR3 parallax bias differ from DR2, and what do these differences reveal about underlying systematics?
• RQ4 What are the limitations of the proposed bias model, and what pathes exist for future refinement?

Key findings

• The parallax bias in Gaia EDR3 depends in a non-trivial way on magnitude, colour, and ecliptic latitude, with different dependencies for five- and six-parameter solutions.
• A parametric, multi-dimensional bias function Delta Z(G, nu_eff, beta) models the EDR3–DR2 parallax differences and informs Z5 and Z6; 137 non-zero coefficients are retained after a significance-based pruning.
• Quasars provide direct bias estimates at faint magnitudes (G greater than ~14) over a limited nu_eff range, while differential methods (LMC, binaries) extend the mapping to brighter regimes and broader colour ranges.
• The authors offer tentative expressions for Z5 and Z6 and document the main residual systematics, noting region- and colour-dependent features and some non-linearities near color/clamping boundaries.
• The global offset for quasars is around a median parallax of approximately -17 to -21 microarcseconds, with spatial variations tied to the scanning law and ecliptic latitude.
• A caveat is that the correction is not definitive; the paper discusses limitations and suggests directions for future improvement.