Quaia: Gaia-unWISE Quasar Catalog

Updated 12 July 2026

Quaia is an all-sky quasar catalog that integrates Gaia optical data with unWISE infrared measurements, using a k-nearest neighbors model to refine redshift estimates.
It offers a large, homogeneous sample with rigorously modeled selection functions, enabling precise cosmological analyses such as BAO, large-scale structure, and primordial non-Gaussianity studies.
The catalog’s design facilitates multi-wavelength extensions and astrometric applications, supporting investigations from radio cross-matching to measurements of Solar System acceleration.

Quaia is the Gaia-unWISE Quasar Catalog, a new, all-sky quasar catalog that samples the largest comoving volume of any existing spectroscopic quasar sample. It combines Gaia quasar candidates with unWISE infrared data, improves redshifts with a $k$ -nearest neighbors model trained on SDSS redshifts, and provides rigorous selection function models; the final catalog has 1,295,502 quasars with $G<20.5$ , and 755,850 candidates in an even cleaner $G<20.0$ sample (Storey-Fisher et al., 2023). The catalog was designed for cosmological and astrophysical quasar studies, especially large-scale structure analyses, and is publicly available at https://zenodo.org/records/10403370 (Storey-Fisher et al., 2023).

1. Scientific context and motivation

Quaia was constructed to exploit the complementarity between Gaia DR3 and unWISE. Gaia supplies highly precise positions, proper motions, and optical photometry, while unWISE contributes mid-infrared measurements that are effective for separating quasars from stars, particularly where optical colors overlap. The resulting aim was a vast, homogeneous, all-sky quasar catalog suited to measurements of large-scale structure, Baryon Acoustic Oscillations, and primordial non-Gaussianity $f_{\rm NL}$ (Storey-Fisher et al., 2023).

The immediate starting point was Gaia’s very large parent sample of quasar candidates with redshift estimates from low-resolution BP/RP spectra. That parent sample was highly homogeneous and complete but had low purity, and 18% of even the bright ( $G<20.0$ ) confirmed quasars had discrepant redshift estimates, defined by $|\Delta z/(1+z)|>0.2$ , relative to the Sloan Digital Sky Survey. Quaia was therefore conceived as a catalog-level refinement of Gaia’s candidate list rather than as a de novo quasar search (Storey-Fisher et al., 2023).

Quaia also sits within a broader Gaia-plus-infrared lineage. Earlier work on Gaia EDR3 and AllWISE demonstrated that mid-infrared color selection and astrometric filtering can produce quasar candidate catalogs with about 80% completeness and about 90% purity, while also showing that purity degrades in crowded fields and in regions with fewer WISE observations (Wu et al., 2021). In parallel, Gaia plus unWISE modeling had already proven useful for lensed-quasar discovery, recovering known lenses with simple WISE–Gaia color cuts that removed $\sim 80$ per cent of previously followed-up contaminants (Lemon et al., 2018). Quaia extended this logic from candidate selection to a cosmology-oriented all-sky catalog with redshift calibration and explicit selection-function modeling.

2. Construction and sample definition

The construction procedure begins from Gaia quasar candidates and cross-matches them to unWISE infrared data. Quaia then applies cuts based on proper motions and Gaia and unWISE colors, reducing the number of contaminants by $\sim 4\times$ . Redshifts are re-estimated with a $k$ -nearest neighbors model trained on SDSS redshifts. On the $G<20.0$ sample, the resulting estimates have only 6% catastrophic errors with $G<20.5$ 0 and 10% catastrophic errors with $G<20.5$ 1, which is a reduction of $G<20.5$ 2 and $G<20.5$ 3, respectively, compared to the Gaia redshifts (Storey-Fisher et al., 2023).

A useful way to view Quaia is as a hierarchy of related samples and derivatives:

Sample	Definition	Size
Parent Gaia candidates	Gaia quasar candidates with BP/RP redshift estimates	6,649,162
Quaia final	Final catalog with $G<20.5$ 4	1,295,502
Quaia cleaner sample	Even cleaner sample with $G<20.5$ 5	755,850
Quaia-VLASS	Cross-matched Quaia–VLASS catalog	43,650

The final catalog is all-sky in construction, and later analyses use different masks and selection thresholds depending on the scientific application. This distinction matters: the catalog itself is full-sky, but the effective footprint used in any downstream measurement is analysis-dependent.

A recurrent point of terminology is that Quaia is introduced as an all-sky spectroscopic quasar sample, because the parent Gaia candidates carry redshift estimates from low-resolution BP/RP spectra, while later cosmological analyses repeatedly treat the catalog as a photometric or spectrophotometric quasar sample because the operational redshifts are the $G<20.5$ 6NN-refined estimates with explicitly modeled uncertainties. This suggests that Quaia is best understood as a hybrid catalog whose utility derives less from a single labeling convention than from the combination of homogeneous selection, calibrated redshift estimation, and a rigorous survey response model (Storey-Fisher et al., 2023).

3. Redshift estimation, survey response, and masking

Quaia’s redshift pipeline uses a $G<20.5$ 7-nearest neighbors model trained on SDSS spectroscopic quasars. In later work, the catalog is described as having spectrophotometric redshifts, and the uncertainties are propagated directly into theory predictions and mock generation. One cosmological analysis reports for $G<20.5$ 8 that the mean redshift uncertainty is $G<20.5$ 9, and constructs redshift distributions as a sum of Gaussians, one per quasar, with associated $G<20.0$ 0 values to account for leakage between tomographic bins (Bermejo-Climent et al., 23 Jan 2026).

Selection-function modeling is central to Quaia’s design. A selection function map in HEALPix format is provided, encoding the probability that a source in each sky pixel enters the catalog. In cosmological applications this map is used to construct overdensity fields of the form

$G<20.0$ 1

where $G<20.0$ 2 is the selection function in pixel $G<20.0$ 3 (Alonso et al., 2023). Subsequent analyses model the selection function with Gaussian processes to account for Milky Way dust extinction, stellar crowding, the Gaia scanning law, and depth variations (Alonso et al., 2023).

Because Quaia is intended for ultra-large-scale analyses, masking strategy is part of the catalog’s practical definition. Different studies impose different conservative cuts. The angular clustering and CMB-lensing study used footprints of 74% of the sky in the low- $G<20.0$ 4 bin and 67% in the high- $G<20.0$ 5 bin after masking regions of low completeness (Alonso et al., 2023). The angular redshift fluctuations analysis applied a selection-function threshold $G<20.0$ 6, producing an effective footprint of about 60% of the sky (Bermejo-Climent et al., 23 Jan 2026). The field-level inference application used a sky coverage of $G<20.0$ 7, corresponding to $G<20.0$ 8, after masking around the Galactic plane and Magellanic Clouds (Andrews et al., 2 Feb 2026). These varying footprints are not inconsistencies in the catalog itself; they are consequences of different robustness requirements.

Tomographic binning is similarly analysis-specific. Quaia has been split into two broad redshift bins at the median redshift $G<20.0$ 9, with mean redshifts near $f_{\rm NL}$ 0 and $f_{\rm NL}$ 1 for large-scale structure studies (Alonso et al., 2023). Other analyses used three bins centered at $f_{\rm NL}$ 2 to reconstruct $f_{\rm NL}$ 3 (Piccirilli et al., 2024), four bins spanning $f_{\rm NL}$ 4 for real-space clustering and duty-cycle measurements (Mascarell et al., 21 Nov 2025), and eight radial bins of width $f_{\rm NL}$ 5 for field-level inference (Andrews et al., 2 Feb 2026). The breadth of these choices reflects the catalog’s unusually wide redshift coverage and the availability of a survey response model that supports multiple compression schemes.

4. Comparative position among quasar catalogs

Quaia’s defining comparative feature is effective volume. The original presentation emphasizes that it samples the largest comoving volume of any existing spectroscopic quasar sample and that its large effective volume makes it highly competitive for cosmological large-scale structure analyses (Storey-Fisher et al., 2023). Later work describes it as about an order of magnitude larger than previous quasar samples and highlights its all-sky homogeneity relative to SDSS, which covers only a fraction of the sky (Storey-Fisher et al., 2023).

This volume is paired with a selection function that is unusually explicit for an all-sky quasar catalog. Later cosmological analyses repeatedly rely on that feature: the catalog is described as having an exceptionally well-defined selection function, and its combination of wide redshift reach and controlled systematics is used to argue that Quaia can probe scales and epochs highly complementary to cosmic shear and to most galaxy clustering samples (Alonso et al., 2023). The same broad sky coverage that aids cosmology also supports applications in astrometry, where reference-frame stability and proper-motion systematics require large homogeneous samples.

A common misconception is to compare Quaia to earlier Gaia-plus-WISE catalogs solely in terms of raw source counts. Earlier catalogs based on astrometric and mid-infrared methods could achieve high purity and substantial completeness, but they were not introduced with the same emphasis on redshift refinement, tomographic analysis, and rigorous cosmological selection functions (Wu et al., 2021). Quaia’s distinctive role lies in integrating these ingredients into a catalog engineered for both astrophysical and cosmological inference.

5. Cosmological analyses enabled by Quaia

Quaia rapidly became a platform for projected large-scale structure cosmology. Using the angular clustering of Quaia and its cross-correlation with Planck CMB lensing, one study obtained $f_{\rm NL}$ 6 and $f_{\rm NL}$ 7, arguing that the catalog’s combination of volume and redshift precision allows the usual degeneracy between $f_{\rm NL}$ 8 and $f_{\rm NL}$ 9 to be broken (Alonso et al., 2023). That analysis also identified a specific systematic concern: contamination of high-redshift quasar–CMB-lensing cross-correlations by extragalactic foregrounds in the Planck lensing map, especially beyond $G<20.0$ 0 (Alonso et al., 2023). A later analysis combining Quaia with ACT DR6 and Planck PR4 reported no evidence for such contamination in the lensing maps, obtained $G<20.0$ 1 from the joint quasar auto- and cross-correlations plus BOSS BAO data, and found that adding Quaia cross-correlations improved the constraint on $G<20.0$ 2 by 12% relative to the lensing auto-spectrum alone, yielding $G<20.0$ 3 (Villagra et al., 11 Jul 2025).

Quaia has also been used to reconstruct the growth history of matter fluctuations out to high redshift. A three-bin analysis measured $G<20.0$ 4, describing this as one of the highest-redshift measurements of $G<20.0$ 5, and reported bias values similar to those of other quasar samples but with a less steep evolution at high redshifts (Piccirilli et al., 2024). An independent clustering analysis over $G<20.0$ 6 found a steady increase in the correlation length, with $G<20.0$ 7 at $G<20.0$ 8, $G<20.0$ 9 at $|\Delta z/(1+z)|>0.2$ 0, $|\Delta z/(1+z)|>0.2$ 1 at $|\Delta z/(1+z)|>0.2$ 2, and $|\Delta z/(1+z)|>0.2$ 3 at $|\Delta z/(1+z)|>0.2$ 4, with slopes consistent with $|\Delta z/(1+z)|>0.2$ 5. Interpreting these measurements with a bias–halo mass relation and a step-function halo occupation model led to characteristic minimum halo masses of quasar hosts of $|\Delta z/(1+z)|>0.2$ 6 across all redshifts and duty cycles rising from $|\Delta z/(1+z)|>0.2$ 7 to $|\Delta z/(1+z)|>0.2$ 8, corresponding to integrated quasar lifetimes of $|\Delta z/(1+z)|>0.2$ 9 years (Mascarell et al., 21 Nov 2025).

The catalog’s large-scale reach also supports measurements of primordial physics. A projected clustering analysis targeting the scale-dependent bias from local-type primordial non-Gaussianity obtained $\sim 80$ 0 at 68% confidence level when combining quasar auto-correlations and cross-correlations with CMB lensing in two tomographic bins under the universality relation $\sim 80$ 1, and $\sim 80$ 2 using the CMB-lensing cross-correlations alone (Fabbian et al., 29 Apr 2025). The same dataset later supported the introduction of angular redshift fluctuations (ARF) as an additional observable,

$\sim 80$ 3

which was combined with projected density and Planck PR4 CMB lensing to obtain $\sim 80$ 4 at 68% confidence level. That result was described as the second tightest constraint on $\sim 80$ 5 using large-scale-structure two-point statistics to date and the best measurement achieved using two-point projected summary statistics, improving by $\sim 80$ 6 on the previous Quaia result (Bermejo-Climent et al., 23 Jan 2026).

Quaia has additionally been used to detect the turnover of the matter power spectrum through its projected clustering and cross-correlation with CMB lensing. In that application, the turnover was detected with a significance of between $\sim 80$ 7 and $\sim 80$ 8, the equality scale was measured at the $\sim 80$ 9 level, and the inferred value was $\sim 4\times$ 0 (Alonso et al., 2024). By combining the turnover measurement with supernova information, the same analysis reported $\sim 4\times$ 1, $\sim 4\times$ 2, and, assuming the COBE-FIRAS temperature, $\sim 4\times$ 3 (Alonso et al., 2024).

The catalog has now also entered field-level cosmology. Applying BORG to the Quaia Clean and Quaia Deep samples, one study reconstructed initial conditions and present-day matter fields over a comoving volume of $\sim 4\times$ 4 with a maximum spatial resolution of $\sim 4\times$ 5, describing this as the largest field-level reconstruction of the observable Universe in terms of comoving volume to date. Cross-correlation of the inferred density field with Planck CMB lensing yielded a detection at $\sim 4\times$ 6 significance (Andrews et al., 2 Feb 2026).

6. Derived catalogs, multi-wavelength extensions, and astrometric applications

Quaia has been extended beyond optical–infrared selection into radio, mock-catalog, and astrometric domains. The Quaia–VLASS cross-match combined the optical catalog of about 1.3 million quasars with 1.9 million VLASS radio sources, adopted a matching radius of $\sim 4\times$ 7, and found contamination from false matches below $\sim 4\times$ 8. The public value-added catalog contains 43,650 sources and supports radio-loudness statistics in which the radio-loud fraction is in good agreement with previous work, below 10%, with no significant large-scale pattern in radio-loudness across the sky (Arsenov et al., 2024).

For cosmological inference and covariance estimation, Quaia now has dedicated mock realizations. One study produced 100 full-sky quasar spectrophotometric mock catalogs with smooth redshift evolution from $\sim 4\times$ 9 to $k$ 0, tailored to analyze Quaia. These mocks combine Augmented Lagrangian Perturbation Theory on the lightcone through the WebON code with the Hicobian hierarchical nonlocal nonlinear bias scheme, and then inject spectrophotometric redshift uncertainties, the angular selection function, and the observed redshift number counts. The resulting catalogs were reported to match the real Quaia maps, redshift uncertainty distributions, $k$ 1, angular power spectra, normalized covariance matrices, and angular two-point correlation functions with excellent agreement (Sinigaglia et al., 19 Sep 2025).

Quaia is also useful in astrometry because quasars define an effectively inertial reference frame. A proper-motion analysis of Quaia quasars was used to measure the Solar System’s acceleration with respect to the quasar rest frame. Using the brighter $k$ 2 subset and simulation-based inference that marginalizes over higher multipoles, the best estimate based on Quaia was

$k$ 3

corresponding to an amplitude of $k$ 4. The same analysis found no significant dependence of the acceleration on source redshift, which was taken as further evidence for a kinematic origin (Nguyen et al., 18 May 2026).

Taken together, these extensions show that Quaia is not only a catalog of quasars but also a survey infrastructure. Its distinctive combination of all-sky homogeneity, explicit selection-function modeling, calibrated redshift uncertainties, and public release has made it a common substrate for projected clustering, cross-correlation cosmology, field-level inference, radio studies, and precision astrometry (Storey-Fisher et al., 2023).