The Sloan Digital Sky Survey Quasar Catalog: sixteenth data release (DR16Q)


In the tradition of previous SDSS data releases, DR16 is accompanied by a quasar catalog. This catalog, which is the largest quasar catalog to date, contains data for 750,414 quasars, of which 225,082 are new. Content in the catalog is explained in detail in Lyke et al. (2020), but salient algorithms are summarized below.

Catalog Contents

The DR16 quasar value-added catalog is contained within two primary FITS tables, DR16Q_Superset_v3.fits (2.4 GB, FITS format) and DR16Q_v4.fits (1.7 GB, FITS format). The column descriptions can be found in the data models.

A superset of observations (DR16Q_Superset_v3.fits) was created by selecting spectra from the final eBOSS SDSS reduction pipeline (v5_13_0). These observations were selected based on the targeting bits assigned for quasar programs outlined in Table 1 of Pâris et al. (2018). Added to this data set were serendipitous quasars previously inspected, additional DR12Q quasars, and DR7Q quasars. It contains ~1.44m observations of quasars, stars, and galaxies. Both catalogs contain data from each of the four observation epochs: SDSS-I/II/III and IV. It is used as the input to the large-scale structure catalogs.

A catalog of quasars, DR16Q_v4.fits, was generated from the superset catalog and contains additional data such as absolute i-band magnitudes and multi-wavelength survey data cross-matched from GALEX, UKIDSS, WISE (W1 and W2), FIRST, 2MASS, 2RXS, XMM-Newton, and Gaia. This quasar-only catalog contains ~750,000 quasars.

The procedures for producing both catalogs are described below.

Superset Catalog Production


Two distinct redshift ranges for quasars were identified as critical to the goals of eBOSS (Dawson et al. 2016). Over a sky area of 7500 sq. deg., the eBOSS mission planned to observe more than 500,000 confirmed quasars in the range 0.8 < z < 2.2 and observe ~120,000 quasars for Lyα forest measurements in the range z > 2.1 (Myers et al. 2015). DR16Q contains >480,000 quasars in the first redshift range and >239,000 quasars for the Lyα forest measurements. To ensure greater completeness and purity, a subsample of spectra were selected for visual inspection and confirmation/correction of their classification and redshifts. This is summarized below.

Visual Inspection Selection Criteria

As in DR14Q, the volume of potential quasar targets in DR16Q was too large to inspect fully. A modified form of the DR14Q algorithm was used, leveraging a new tool, QuasarNET, to identify the most likely candidates the pipeline would classify incorrectly.

The first step was to re-apply the DR14Q algorithm to all eBOSS observations. This process is as follows:

We start with the first five identifications produced by the SDSS spectroscopic pipeline (i.e. identifications corresponding to the five lowest values of reduced-χ2), the pipeline redshifts (zpipeline), and the ZWARNING: Warnings for SDSS spectra quality flag. We then apply the following algorithm:

1. If the best model fit for the observation (first SDSS pipeline identification) is STAR, then the resulting classification is STAR;

2. If the best model fit is GALAXY with zpipeline < 1, then the resulting classification is GALAXY;

3. If the best model fit is GALAXY with zpipeline ≥ 1 and at least one other SDSS pipeline identification is GALAXY, then the resulting classification is GALAXY;

4. If the best model fit is QSO and 2 or more model fits are STAR, the observation is classified as a STAR;

5. If the best model fit is QSO, fewer than two model fits are STAR, and ZWARNING = 0, the object is classified as a QSO;

6. If an observation meets none of these criteria it is selected for visual inspection (VI).

After this initial classification, QuasarNET was applied to the data. VI-flagged objects it classified as quasars were reclassified as QSO. This combined approach flagged < 1% of the objects for VI. While a small percentage, this still represented > 8000 spectra.

Visual Inspection Procedure

As in DR14Q, objects selected for visual inspection were reclassified and quasar redshifts were corrected where needed. Spectra that were not quasars upon visual inspection, or had a very low signal-to-noise ratio were classified as “non-quasars” (marked as GALAXY) and did not have redshifts corrected.

Unlike the SDSS-III/BOSS quasar catalogs, features of interest such as broad absorption line quasars (BALs) and damped Lyα systems (DLAs) were not flagged. The small percentage of visual inspections did not constitute a large enough sample for these featured objects in comparison to a pipeline. Instead, these features were found via automated algorithms. The BAL algorithm is adapted from Guo et al. (2018) and the DLA algorithm from Chabanier et al. (2018). Full details can be found in Lyke et al. (2020).

Quasar-Only Catalog Production (DR16Q)

Duplicate Removal

As some quasar targets were part of multi-epoch observation campaigns (TDSS; RM) the selection of quasars relied on visual inspection data and a spectral signal-to-noise ratio to determine the “primary record.” Where visual inspection data was available for a quasar, the observation with the highest confidence was selected. In the absence of VI data the primary observation selected was that which had the highest SN_MEDIAN_ALL value.

Final Quasar Selection

The final classification of whether an object was a quasar relied on visual inspection data, confident pipeline classifications, and previous catalog classifications. For each object a confident visual inspection would override an automated classification, but a low-confidence inspection would not. The status of each object was recorded in IS_QSO_FINAL in the superset and the quasar-only catalog. Any object with IS_QSO_FINAL > 0 was selected as a quasar for DR16Q.

Available Redshifts and Selection of a “Primary” Redshift

DR16Q offers more options for redshift estimates than previous catalog releases. The “primary” redshift estimate is recorded in the column, Z, with the source for this redshift recorded in SOURCE_Z. Only visual inspection and the SDSS automated pipeline redshift were used to populate these columns; PCA systemic redshifts, emission line redshifts, and QuasarNET redshifts are included in their own columns, but were not used to populate Z.

DR16Q also includes two options for the “Hewett and Wild” improved redshifts taken from Hewett and Wild (2010), based on DR6 data, and Shen et al. (2011), based on DR7Q data. The former are presented in the column, Z_DR6Q_HW, and the latter in Z_DR7Q_HW.

Previously released SDSS quasar catalogs

We recommend this catalog, DR16Q, is used for identifying confirmed quasars, as this catalog corrects some information in previous catalogs. See Appendix C in Lyke et al. (2020) for a list of known bugs in those catalogs. Previously released SDSS quasar catalogs are still available on the SAS, and are described in:

Pâris et al. (2018) for DR14Q, see also here;
Pâris et al. (2017) for DR12Q;
Pâris et al. (2014) for DR10Q;
Pâris et al. (2012) for DR9Q.

Corresponding data models for these quasar catalogs can be found here.