The Seventeenth Data Release of the Sloan Digital Sky Surveys: Complete Release of MaNGA, MaStar, and APOGEE-2 Data

Nidever et al. (2015) describe the original reduction procedure for APOGEE data, and the various APOGEE Data Release papers present updates (Abolfathi et al. 2018; Holtzman et al. 2018; Jönsson et al. 2020; J. Holtzman et al. 2022, in preparation). For DR17, at the visit reduction level, a small change was made to the criteria by which pixels are flagged as being potentially affected by poor sky subtraction.

The routines for the combination of the individual visit-spectra were rewritten for DR17 to incorporate a new radial-velocity analysis, called Doppler (Nidever et al. 2021). Doppler performs a least squares fit to a set of visit-spectra, solving simultaneously for basic stellar parameters (T_eff, $\mathrm{log}\,g$, and [M/H]) and the radial velocity (RV) for each visit. The fitting is accomplished by using a series of Cannon (Ness et al. 2015; Casey et al. 2016) models to generate spectra for arbitrary choices of stellar parameters across the Hertzsprung-Russell diagram (from 3500 K to 20,000 K in T_eff); the Cannon models were trained on a grid of spectra produced using Synspec (e.g., Hubeny & Lanz 2017; Hubeny et al. 2021) with Kurucz model atmospheres (Kurucz 1979; Castelli & Kurucz 2003; Munari et al. 2005). The primary outputs of Doppler are the RVs; while the stellar parameters from Doppler are stored, they are not adopted as the final values (see "ASPCAP," Section 4.2.2 below). The Doppler routine produces slightly better results for RVs in most cases, as judged by scatter across repeated visits of stars. Details will be given in J. Holtzman et al. (2022, in preparation), but, for example, for ∼85,000 stars that have more than three visits, VSCATTER< 1 km s⁻¹, TEFF< 6000 K, and no additional data since DR17; the median VSCATTER is reduced from 128 to 96 m s⁻¹.

In addition to the new methodology, the RVs for faint stars were improved. This was accomplished by making an initial combination of the visit-spectra using only the barycentric correction. This initial combination provided a combined spectrum from which an RV was determined. The RV for each individual visit was then determined separately, but was required to be within 50 km s⁻¹ of the original estimate. This yielded a higher fraction of successful RVs for faint stars, as judged by looking at targets in the nearby dwarf spheroidal galaxies.

Stellar parameters and abundances are determined using the APOGEE Stellar Parameters and Chemical Abundance Pipeline (ASPCAP; García Pérez et al. 2016) that relies on the FERRE optimization code (Allende Prieto et al. 2006). ¹⁵⁸

The basic methodology of ASPCAP remained the same for DR17 as in previous releases, but new synthetic spectral grids were created. These took advantage of the new, non-local thermodynamic equilibrium (NLTE) population calculations by Osorio et al. (2020) for four elements: Na, Mg, K, and Ca; as discussed in Osorio et al. (2020), the H-band abundance differences between LTE and NLTE were always less than 0.1 dex. Adopting these calculations, however, required the adoption of a different spectral synthesis code from that used in the last several APOGEE data releases: for DR17, the Synspec code (e.g., Hubeny & Lanz 2017; Hubeny et al. 2021) was adopted for the primary analysis instead of the Turbospectrum code (Alvarez & Plez 1998; Plez 2012) used in previous releases. This was not a straightforward choice because, while Synspec allows the NLTE levels to be used, it calculates the synthetic spectra under the assumption of plane parallel geometry, which becomes less valid for the largest giant stars. On the other hand, the Turbospectrum can use spherical geometry, but does not accommodate NLTE populations to be specified.

DR17 uses multiple subgrids to span from T_eff = 3000 K (M dwarf) to T_eff = 20,000 K (BA), with $\mathrm{log}\,g$ ranges from 0 to 5 (3 to 5 for the BA grid). The full details of these grids and the reliability of the parameters as a function of stellar type are provided in J. Holtzman et al. (2022, in preparation). Modifications to the linelists used for the syntheses are described in Smith et al. (2021), which is an augmentation to prior linelist work for APOGEE (Shetrone et al. 2015; Hasselquist et al. 2016; Cunha et al. 2017).

The ASPCAP results from the new Synspec grid are the primary APOGEE DR17 results, and the majority of users will likely be satisfied with the results in this catalog; only this primary catalog will be loaded into the CAS. However, unlike prior releases, DR17 also includes supplemental analyses constructed using alternate libraries that have different underlying physical assumptions. The different analyses in DR17 are provided in separate summary files and include the following:

• 1.  
  the primary library using Synspec including NLTE calculations for Na, Mg, K, and Ca (with files on the SAS under dr17/synspec_rev1); ¹⁵⁹
• 2.  
  one created using Synspec, but assuming LTE for all elements (files under dr17/synspec_lte);
• 3.  
  another created using Turbospectrum 20 (files under dr17/turbo20), using spherical geometry for $\mathrm{log}g\lt =3$;
• 4.  
  one created with Turbospectrum, but with plane parallel geometry (files under dr17/turbo20_pp) for all stars.

All of the libraries use the same underlying MARCS stellar atmospheres for stars with T_eff < 8000 K, computed with spherical geometry for $\mathrm{log}g\lt =3$. A full description of these spectral grids will be presented in J. Holtzman et al. (2022, in preparation), and a focused discussion on the differences between the libraries and the physical implications will be presented in Y. Osorio et al. (2022, in preparation). In summary, however, the differences are subtle in most cases. We encourage those using the APOGEE DR17 results to clearly specify the catalog version that they are using in their analyses. ¹⁶⁰

For DR17, we present 20 elemental abundances: C, C i, N, O, Na, Mg, Al, Si, S, K, Ca, Ti, Ti ii, V, Cr, Mn, Fe, Co, Ni, and Ce. In DR16, we attempted to measure the abundances of Ge, Rb, and Yb, but given the poor results for extremely weak lines, we did not attempt these in DR17. While we attempted measurements of P, Cu, Nd, and ¹³C in DR17, these were judged to be unsuccessful. Overall, the spectral windows used to measure the abundances were largely unchanged, but several additional windows were added for cerium, such that the results for Ce appear to be significantly improved over those in DR16.

As in DR16, both the raw spectroscopic stellar parameters as well as the calibrated parameters and abundances are provided. Calibrated-effective temperatures are determined by a comparison to photometric-effective temperatures, as determined from the relations of González Hernández & Bonifacio (2009), using stars with low reddening. Calibrated surface gravities are provided by comparison to a set of surface gravities from asteroseismology (Serenelli et al. 2017; M. Pinsonneault et al. 2022, in preparation) and isochrones (Berger et al. 2020). For DR17, the surface gravity calibration was applied using a neural network, unlike previous data releases where separate calibrations were derived and applied for different groups (red giants, red clump, and main sequence) of stars. The new approach eliminates small discontinuities that were previously apparent, and is described in more detail in J. Holtzman et al. (2022, in preparation). For the elemental abundances, calibration just consists of a zero-point offset (separately for dwarfs and giants), determined by setting the median abundance [X/M] of solar metallicity stars in the solar neighborhood with thin disk kinematics such that [X/M] = 0.

Additional details on the ASPCAP changes are described in J. Holtzman et al. (2022, in preparation).

Several other modifications were made for DR17.

• 1.  
  The summary data files for APOGEE that are available on the SAS now include data from the Gaia Early Data Release 3 (EDR3) for the APOGEE targets (Gaia Collaboration et al. 2016, 2021). Positional matches were performed by the APOGEE team. More specifically, the following data are included:

  • (a)  
    Gaia EDR3 identifiers (Gaia Collaboration et al. 2021),
  • (b)  
    Gaia EDR3 parallaxes and proper motions (Lindegren et al. 2021),
  • (c)  
    Gaia EDR3 photometry (Riello et al. 2021),
  • (d)  
    Gaia EDR3 radial-velocity (RV) measurements (Seabroke et al. 2021),
  • (e)  
    Distances and uncertainties following Bailer-Jones et al. (2021).
• 2.  
  Likely membership for a set of open clusters, GCs, and dwarf spheroidal galaxies, as determined from position, RV, proper motion, and distance, is provided in a MEMBERS column. More specifically, initial memberships were computed based on position and the literature RVs, and these are then used to determine the proper-motion (PM) and distance criteria. Literature RVs were taken from the following:

  • (a)  
    APOGEE-based mean RVs for the well-sampled "calibration clusters" in Holtzman et al. (2018),
  • (b)  
    mean RVs for GCs from Harris (2010), ¹⁶¹
  • (c)  
    mean RVs for dwarf spheroidal galaxies from McConnachie (2012).

  Users interested in the properties of the clusters or the satellite galaxies are encouraged to do more detailed membership characterization and probabilities (e.g., Masseron et al. 2019; Mészáros et al. 2020; Hasselquist et al. 2021; R. Schiavon et al. 2022, in preparation; M. Shetrone et al. 2022, in preparation)
• 3.  
  Some spectroscopic binary identification is provided through bits in the STARFLAG and ASPCAPFLAG bitmasks. A more comprehensive analysis of spectroscopic binaries is provided in a VAC (see Section 4.4.1 below).

We encourage those utilizing these data in our summary catalogs to cite the original references as given above.

The first set of APOGEE VACs describes special categories of objects in APOGEE data and in most cases provides additional information/characteristics for these objects. They are as follows:

• 1.  
  Open Cluster Chemical Abundances and Mapping Catalog (OCCAM): The goal of OCCAM is to leverage the APOGEE survey to create a large, uniform catalog of open cluster chemical abundances and use these clusters to study Galactic-chemical evolution. The catalog contains the average chemical abundances for each cluster and the membership probability estimates for APOGEE stars in the cluster area. We combine PM and RV measurements from Gaia EDR3 (Gaia Collaboration et al. 2021) with RV and metallicity measurements from APOGEE to establish the cluster membership probabilities for each star observed by APOGEE. The VAC includes 26,699 stars in the areas of 153 cataloged disk clusters. Detailed descriptions of the OCCAM survey, including targeting and the methodology for membership determinations, are presented in Frinchaboy et al. (2013), Donor et al. (2018), and Donor et al. (2020). This third catalog from the OCCAM survey includes 44 new open clusters, including many in the southern hemisphere and those targeted specifically in GC-size (R_GC ) ranges with little coverage in the DR16 catalog (specific targeting described in Beaton et al. 2021; Santana et al. 2021). Average RV, PM, and abundances for reliable ASPCAP elements are provided for each cluster, along with the visual quality determination. Membership probabilities based individually upon PM, RV, and [Fe/H] are provided for each star; stars are considered 3σ members if they have probability >0.01 in all three membership dimensions. ¹⁶⁴ The results and caveats from this VAC will be discussed thoroughly in N. Myers et al. (2022, in preparation).
• 2.  
  APOGEE Red-clump (RC) Catalog: DR17 contains an updated version of the APOGEE red-clump (APOGEE-RC) catalog. This catalog is created in the same way as the previous DR14 and DR16 versions of the catalog, with a more stringent $\mathrm{log}g$ cut compared to the original version of the catalog (Bovy et al. 2014). The catalog contains 50,837 unique stars, about 30% more than in DR16. The catalog is created using the spectrophotometric technique first presented in Bovy et al. (2014) that results in a rather pure sample of red-clump stars (e.g., minimal contamination from red-giant-branch, secondary-red-clump, and asymptotic-giant-branch stars that have similar CMD and H-R positions). Bovy et al. (2014) estimated a purity of ∼95%. The narrowness of the RC locus in color-metallicity-luminosity space allows distances to the stars to be assigned with an accuracy of 5%–10%, which exceeds the precision of spectrophotometric distances in other parts of the H-R diagram. We recommend users adopt the most recent catalog (DR17) for their analyses; additional discussion on how to use the catalog is given in Bovy et al. (2014). While the overall data model is similar to previous versions of the catalog, the proper motions are from Gaia EDR3 (Gaia Collaboration et al. 2021; Lindegren et al. 2021).
• 3.  
  APOGEE-Joker: The APOGEE-Joker VAC contains posterior samples for the binary-star orbital parameters (Keplerian orbital elements) for 358,350 sources with three or more APOGEE visit-spectra that pass a set of quality cuts as described in A. Price-Whelan et al. (2022, in preparation). The posterior samples are generated using The Joker, a custom Monte Carlo sampler designed to handle the multi-modal likelihood functions that arise when inferring orbital parameters with sparsely sampled or noisy RV time data (Price-Whelan et al. 2017). This VAC deprecates the previous iterations of the catalog (Price-Whelan et al. 2018, 2020). For 2819 stars, the orbital parameters are well constrained, and the returned samples are effectively unimodal in period. For these cases, we use the sample(s) returned from The Joker to initialize standard MCMC sampling of the Keplerian parameters using the time-series optimized MCMC code known as exoplanet ¹⁶⁵ (Foreman-Mackey et al. 2021) and provide these MCMC samples. For all stars, we provide a catalog containing metadata about the samplings, such as the maximum a posteriori (MAP) parameter values and sample statistics for the MAP sample. A. Price-Whelan et al. (2022, in preparation) describe the data analysis procedure in more detail, and define and analyze a catalog of ≳40,000 binary-star systems selected using the raw orbital parameter samples released in this VAC.
• 4.  
  Double-lined Spectroscopic Binaries in APOGEE Spectra: Generally, APOGEE fibers capture a spectrum of single stars. Sometimes, however, there may be multiple stars of comparable brightness with the sky separations closer than the fiber radius whose individual spectra are captured by the same recorded spectrum. Most often, these stars are double-lined spectroscopic binaries (SBs) or higher-order multiples, but on an occasion they may also be chance line-of-sight alignments of random field stars (most often observed toward the Galactic center). Through analyzing the cross-correlation function (CCF) of the APOGEE spectra, Kounkel et al. (2021) have developed a routine to automatically identify these SBs using Gaussian deconvolution of the CCFs (Kounkel 2021), ¹⁶⁶ and to measure RVs of the individual stars. The catalog of these sources and the subcomponent RVs are presented here as a VAC. For the subset of sources that had a sufficient number of measurements to fully characterize the motion of both stars, the orbit is also constructed. The data obtained though 2020 April/May were processed with the DR16 version of the APOGEE RV pipeline, and this processing was made available internally to the collaboration as an intermediate data release. All of the SBs identified in this internal data release have undergone rigorous visual vetting to ensure that every component that can be detected is included and that spurious detections have been removed. However, the final DR17 RV pipeline is distinct from that used for DR16 (summarized above; J. Holtzman et al. 2022, in preparation), and the reductions are sufficiently different that they introduce minor discrepancies within the catalog. In comparison to DR16, the DR17 pipeline limits the span of the CCF for some stars to a velocity range around the mean RV to ensure a more stable overall set of RV measurements; on the other hand, the DR16 pipeline itself may fail on a larger number of individual visit-spectra and thus not produce a full set of outputs. For the sources that have both good parameters and a complete CCF coverage for both DR16 and DR17, the widely resolved components of SBs are generally consistent with one another; close companions that have only small RV separations are not always identified in both data sets. For this reason, SBs that could be identified in both the DR16 and DR17 reductions are kept as separate entries in the catalog. Visual vetting was limited only to the data processed with the DR16 pipeline (e.g., data through 2020 April/May); the full automatic deconvolutions of the DR17 CCFs are presented as-is.

VACs providing distances and other properties (mostly related to orbital parameters) are released (or rereleased) as follows:

• 1.  
  StarHorse Distances, Extinctions, and Stellar Parameters for APOGEE DR17 + Gaia EDR3: We combine the high-resolution spectroscopic data from APOGEE DR17 with the broadband photometric data from 2MASS, unWISE, and PanSTARRS-1, as well as the parallaxes from Gaia EDR3. Using the Bayesian isochrone-fitting code StarHorse (Santiago et al. 2016; Queiroz et al. 2018), we derive distances, extinctions, and astrophysical parameters. We achieve typical distance uncertainties of ∼5% and extinction uncertainties in V-band amount to ∼0.05 mag for stars with available PanSTARRS-1 photometry, and ∼0.17 mag for stars with only infrared photometry. The estimated StarHorse parameters are robust to the changes in the Galactic priors assumed and the corrections for Gaia parallax zero-point offset. This work represents an update of DR16-based results presented in Queiroz et al. (2020).
• 2.  
  APOGEE- astroNN : The APOGEE-astroNN VAC holds the results from applying the astroNN deep-learning code to APOGEE spectra to determine the stellar parameters, individual stellar abundances (Leung & Bovy 2019a), distances (Leung & Bovy 2019b), and ages (Mackereth et al. 2019a). For DR17, we have retrained all of the neural networks using the latest data, i.e., APOGEE DR17 results for the abundances, Gaia EDR3 parallax measurements, and an intermediate APOKASC data set with stellar ages (v6.6.1, 2020 March using DR16 ASPCAP). Additionally, we augmented the APOKASC age data with the low-metallicity asteroseismic ages from Montalbán et al. (2021) to improve the accuracy of ages at low metallicities; the Montalbán et al. (2021) analysis is similar to that of APOKASC, but performed by an independent team. As in DR16, we correct for the systematic differences between spectra taken at LCO and APO by applying the median difference between stars observed at both observatories. In addition to abundances, distances, and ages, properties of the orbits in the Milky Way (and their uncertainties) for all stars are computed using the fast method of Mackereth & Bovy (2018) assuming the MWPotential2014 gravitational potential from Bovy (2015). Typical uncertainties in the parameters are 35 K in T_eff, 0.1 dex in $\mathrm{log}\,g$, 0.05 dex in elemental abundances, 5% in distance, and 30% in age. Orbital properties such as the eccentricity, maximum height above the mid-plane, radial, and vertical action are typically precise to 4%–8%.

A number of different pipelines are available for extracting spectral parameters from the APOGEE spectra. These pipelines generally manage to achieve optimal performance for red giants and, increasingly, G and K dwarfs, which compose the bulk of the stars in the catalog. However, the APOGEE2 catalog contains a number of parameter spaces that are often not well characterized by the primary pipelines. Such parameter spaces include pre-main-sequence stars and low-mass stars, with their measured parameters showing systematic T_eff and $\mathrm{log}g$ deviations making them inconsistent from the isochrones and the main sequence. OBA stars are also less well constrained and in prior data releases many were classified as F dwarfs (due to grid-edge effects) and have their T_eff underestimated in the formal results. By using data-driven techniques, we attempt to fill in those gaps to construct a unified model of APOGEE spectra. In the past, we have developed a neural network, APOGEE Net (Olney et al. 2020), which was shown to perform well to extract T_eff, $\mathrm{log}g$, and [Fe/H] on all stars with T_eff < 6500 K, including pre-main-sequence stars. We now expand these efforts to also characterize hotter stars with 6500 < T_eff < 50,000 K. APOGEE NET II is described in Sprague et al. (2022).

Mock catalogs made by making simulated observations of sophisticated galaxy simulations provide unique opportunities for observational projects, in particular, the ability to test for or constrain the impact of selection functions, field plans, and algorithms on scientific inferences. One of the most realistic galaxy simulations to date is the Latte simulation suite, which uses FIRE-2 (Hopkins et al. 2018) to produce galaxies in Milky Way mass halos in a cosmological framework (Wetzel et al. 2016). Sanderson et al. (2020) translated three of the simulations into realistic mock catalogs (using three solar locations, resulting in nine catalogs), known as the Ananke simulations. ¹⁶⁷ Ananke contains key Gaia measureables for the star particles in the simulations, and these include RV, proper motion, parallax, and photometry in the Gaia bands as well as chemistry (10 chemical elements are tracked in the simulation), and other stellar properties. Because the input physics and the global structure of the model galaxy are known, these mock catalogs provide an experimental laboratory to make connections between the resolved stellar populations and global galaxy studies.

In this VAC, Ananke is expanded to permit APOGEE-style sampling of the mock catalogs. For all observed quantities, both the intrinsic, e.g., error-free, and the observed values are reported; the observed values are the intrinsic values convolved with an error-model derived from observational data for similar object types. As described in Nikakhtar et al. (2021), Ananke mock catalogs now contain the following: (i) 2MASS (JHK_s ) photometry and reddening, (ii) abundance uncertainties following APOGEE DR16 performance (following Jönsson et al. 2020; Poovelil et al. 2020), and (iii) a column that applies a basic survey map (Zasowski et al. 2013, 2017; Beaton et al. 2021; Santana et al. 2021). The full mock catalogs are released such that users can impose their own selection functions to construct mock APOGEE surveys in the simulation. Mock-surveys can then be used to test the performance of methods and algorithms to recover the true underlying galactic physics as demonstrated in Nikakhtar et al. (2021).

In DR17, we use different spectral templates to model the galaxy continuum for emission-line measurements than we use for stellar kinematics measurements. In DR15, we used the same templates in both cases, but as discussed by Law et al. (2021b), these template sets diverged starting with our ninth internal data set (MPL-9; between DR15 and DR17). For the emission-line measurements, the new templates are based on the MaStar survey, allowing us to take advantage of the full MaNGA spectral range (3600–10000 Å) and, e.g., model the [S iii]λ λ 9071,9533 Å doublet and some of the blue Paschen lines. For the stellar kinematics measurements, we have continued to use the same templates used in DR15, the MILES-HC library, taking advantage of its modestly higher spectral resolution than MaStar. Since MILES only spans between 3575 and 7400 Å, this means MaNGA stellar kinematics do not include, e.g., contributions from the calcium NIR triplet near 8600 Å.

In DR17, we provide DAP emission-line measurements based on two different continuum template sets, both based on the MaStar Survey (Yan et al. 2019; and Section 6), and referred to as MASTARSSP and MASTARHC2. There are four different analysis approaches, indicated by DAPTYPE. Three use MASTARSSP, with three different spatial binning approaches, and the fourth uses MASTARHC2.

The template set referred to as the MASTARSSP library by the DAP are a subset of the simple-stellar-population (SSP) models provided by Maraston et al. (2020). Largely to decrease execution time, we down-selected templates from the larger library provided by Maraston et al. (2020) to only those spectra with a Salpeter initial mass function (IMF) and the following grid in SSP age and metallicity, for a total of 54 spectra:

• 1.  
  Age/[1 Gyr] = 0.003, 0.01, 0.03, 0.1, 0.3, 1, 3, 9, 14
• 2.  
  $\mathrm{log}(Z/{Z}_{\odot })=-1.35$, −1., −0.7, −0.33, 0, 0.35.

Extensive testing was done to check differences in stellar-continuum fits based on this choice; the small differences that were found are well within the limits described by Belfiore et al. (2019). Section 5.3 of Law et al. (2021a) shows further analysis, including a direct comparison of results for the BPT emission-line diagnostics plots when using either the MASTARHC2 or MASTARSSP templates, showing that the templates have a limited effect on their analysis. Importantly, note that the DAP places no constraints on how these templates can be combined (e.g., unlike methods that use the Penalized PiXel-Fitting, or pPXF; Cappellari & Emsellem 2004; Cappellari 2017, implementation of regularized weights), and the weight applied to each template is not used to construct luminosity-weighted ages or metallicities for the fitted spectra. The use of the SSP models, as opposed to the spectra of single stars, is meant only to impose a physically relevant prior on the best-fitting continua, even if minimally so compared to more sophisticated stellar-population modeling.

The template set referred to as the MASTARHC2 ¹⁷⁰ library by the DAP is a set of 65 hierarchically clustered templates based on ∼2800 MaStar spectra from MPL-10. Only one of the four DAPTYPEs provided in DR17 uses these templates; however, we note that the results based on these templates are the primary data sets used by Law et al. (2021a, 2021b) to improve the DRP (see above). The approach used to build the MASTARHC2 library is inspired by, but different in many details from, the hierarchical clustering (HC) method used to build the MILESHC library (see Westfall et al. 2019, their Section 5), as described below.

The principles of the HC approach used by Westfall et al. (2019) to construct the MILESHC library are maintained, except we perform the clustering for the MASTARHC2 library in two steps. The first step clusters spectra based on their low-order continuum differences, leading to a set of base clusters. We use pPXF (Cappellari & Emsellem 2004; Cappellari 2017) to perform a least-squares fit of each spectrum using every other spectrum; however, we do not include Gaussian kernel terms or polynomial continuum optimization, meaning the least-squares fit simply optimizes the scaling between the two spectra. We use the rms difference between the best-fit spectra as the clustering distance, and the distance matrix is used to construct eight base clusters. The choice of eight clusters was based on an a qualitative assessment of the appropriate number that separated MaStar spectra into distinct types. The second step uses pPXF to fit each spectrum using every other spectrum within its base cluster. In this step, we modestly degrade the resolution of the template being fit with σ = 1 pixel, and then our pPXF fit includes a freely fit Gaussian kernel with bounds of a ±1 pixel shift and a 0.1–2 pixel broadening. This was done in the same way across all parts of the spectra. We also include a multiplicative Legendre polynomial of order 100 to optimize the continuum match between the two templates. The very high-order fit (the choice of the exact number of 100 was arbitrary) acts like a high-pass filter on the differences between the two spectra, ensuring that the optimized rms difference between the two spectra is driven by the high-order (line) structure differences. The spectra within each base cluster are organized into template clusters and visually inspected. The visual inspection leads to iterations on the number of template clusters in each base cluster, as well as removing some of the spectra from the analysis. The number of template clusters per base cluster ranged from 6 to 16, depending on a by-eye assessment of the spectra in each template cluster. The final assignment of each MaStar spectrum (identified by its MANGAID) to a template cluster is provided in the DAP code repository. ¹⁷¹ Note that 34 of 99 clusters were not included in the MASTARHC2 library because they were either composed of single stars, resulted in noisy spectral stacks, contained isolated specific data-reduction artifacts, or contained a set of spectra that were considered too disparate for a single cluster. For the vetted set of 65 template clusters, the median number of spectra per cluster is 14, but the range is from 2 to more than 300.

With the assignments in hand, we combine the spectra in each template cluster as follows. We first scale each spectrum by their median flux and create an initial stack, weighting each spectrum by its median S/N. We then calculate the ratio of each spectrum to the stacked spectrum and fit this with an order-14 Legendre polynomial, which provides a low-order correction function to the continuum shape of each spectrum. The specific choice of order 14 was driven by a desire to match the choice made in the DAP fitting of galaxy spectra, which was justified in Westfall et al. (2019). We constrain the correction function to be no more than a factor of 2, which is particularly important to the stacks of late-type stars with very little flux toward the blue end of MaStar's spectral range. The low-order correction function is then applied to each spectrum in the template cluster before the final S/N-weighted stack. The error vector for each stack is the quadrature sum of the propagated error from the stacking operations and the (typically much more significant) standard deviation measured for the spectra in the stack. The final spectra in the MASTARHC2 library are shown in Figure 6.

Zoom In Zoom Out Reset image size

In DR17, we have added spectral index measurements that are more suited to stacking analyses and coaddition of spectra among spaxels, such as those based on definitions of Burstein et al. (1984) and Faber et al. (1985). These measurements are particularly useful for low-S/N spaxels.

The motivation for this change emerges from the fact that, in DAP's hybrid binning scheme, the spectral index measurements are performed on individual spaxels, which can have a very low S/N (see Westfall et al. 2019; their Section 9). Westfall et al. (2019) recommend improving the precision of the spectral index measurements using specific aggregation calculations that closely match the results obtained by performing the measurements on stacked spectra over the same spatial regions (specifically, see their Section 10.3.3). However, the comparison between an aggregated index and an index measured using a stacked spectrum is not mathematically identical for the index definitions used by Westfall et al. (2019). Motivated by the analysis of Molina et al. (2020), the DAP calculates the spectral indices (specifically the absorption line indices) using two definitions for DR17: (1) those definitions provided by Worthey (1994) and Trager et al. (1998) and (2) earlier definitions provided by Burstein et al. (1984) and Faber et al. (1985). The advantage of the definitions provided by Burstein et al. (1984) and Faber et al. (1985) is that they allow for a mathematically rigorous aggregation of spectral indices, as we derive below.

Following the derivation by Westfall et al. (2019), we define a utility function, which is a sum of pixel values, multiplied by pixel width,

$$\begin{eqnarray}&&S(y)\equiv {\int }_{{\lambda }_{1}}^{{\lambda }_{2}}y\ d\lambda \approx \sum _{i}{y}_{i}\ {{dp}}_{i}\ d{\lambda }_{i},\end{eqnarray} \tag{ 1 }$$

where y is usually, but not always a function describing the flux in the spectrum, f(λ), and dp_i is the fraction of spectral pixel i (with width dλ_i ) in the passband defined by λ₁ < λ < λ₂. Note that masked pixels in the passband are excluded by setting dp_i = 0, and S(1) = Δλ ≡ λ₂ − λ₁ if no pixels are masked. We can then define a linear continuum between two sidebands, referred to as the blue and red sidebands, as

$$\begin{eqnarray}&&C(\lambda )=(\langle f{\rangle }_{\mathrm{red}}-\langle f{\rangle }_{\mathrm{blue}})\ \displaystyle \frac{\lambda -{\lambda }_{\mathrm{blue}}}{{\lambda }_{\mathrm{red}}-{\lambda }_{\mathrm{blue}}}+\langle f{\rangle }_{\mathrm{blue}},\end{eqnarray} \tag{ 2 }$$

where f is the spectrum flux density, λ_blue and λ_red are the wavelengths at the center of the two sidebands, and 〈f〉 = S(f)/S(1).

The absorption-line index definitions used by Worthey (1994) and Trager et al. (1998) are as follows:

$$\begin{eqnarray}{{ \mathcal I }}_{\mathrm{WT}}=\left\{\begin{array}{ll}S(1-f/C), & {\rm{for}}\,{\rm{\mathring{\rm A} }}\,{\rm{units}}\\ -2.5\mathrm{log}\left[\langle f/C\rangle \right], & {\rm{for}}\,{\rm{magnitude}}\,{\rm{units}},\end{array}\right.\end{eqnarray} \tag{ 3 }$$

where the measurements are made on a rest-wavelength spectrum. ¹⁷² Under this definition, the integration is performed over the ratio of the flux to a linear continuum, which means that the sum of, say, two index measurements is not identical to a single index measurement made using the sum of two spectra. In contrast, Burstein et al. (1984) and Faber et al. (1985) define the following:

$$\begin{eqnarray}{{ \mathcal I }}_{\mathrm{BF}}=\left\{\begin{array}{ll}S(1)-S(f)/{C}_{0}, & {\rm{for}}\,{\rm{\mathring{\rm A} }}\,{\rm{units}}\\ -2.5\mathrm{log}\left[\langle f\rangle /{C}_{0}\right], & {\rm{for}}\,{\rm{magnitude}}\,{\rm{units}},\end{array}\right.\end{eqnarray} \tag{ 4 }$$

where C₀ is the value of the continuum, C(λ), at the center of the main passband. Note that, given that C(λ) is linear and assuming no pixels are masked, S(C) = C₀ Δλ. Using the definition in Equation (4), we can calculate a weighted sum of indices using the value of the continuum, C₀, for each index as the weight to obtain

$$\begin{eqnarray}&&\displaystyle \frac{{\sum }_{i}{C}_{0,i}{{ \mathcal I }}_{\mathrm{BF}}}{{\sum }_{i}{C}_{0,i}}={\rm{\Delta }}\lambda -\displaystyle \frac{{\sum }_{i}S{(f)}_{i}}{{\sum }_{i}{C}_{0,i}},\end{eqnarray} \tag{ 5 }$$

assuming no pixels are masked such that S(1) = Δλ. That is, the weighted sum of the individual indices is mathematically identical (to within the limits of how error affects the construction of the linear continuum) to the index measured for the sum (or mean) of the individual spectra. Similarly, for the indices in magnitude units, we find the following:

$$\begin{eqnarray}&&-2.5\mathrm{log}\left[\displaystyle \frac{{\sum }_{i}{C}_{0,i}{10}^{-0.4{{ \mathcal I }}_{\mathrm{BF}}}}{{\sum }_{i}{C}_{0,i}}\right]=-2.5\mathrm{log}\left[\displaystyle \frac{{\sum }_{i}\langle f{\rangle }_{i}}{{\sum }_{i}{C}_{0,i}}\right].\end{eqnarray} \tag{ 6 }$$

Given the ease with which one can combine indices in the latter definition, we provide both ${{ \mathcal I }}_{\mathrm{BF}}$ (in the SPECINDEX_BF extension of the DAP MAPS file) and C₀ (in SPECINDEX_WGT) for all absorption-line indices in DR17, along with the original definitions (${{ \mathcal I }}_{\mathrm{WT}};$ SPECINDEX) provided in DR15/DR16.

Two MaNGA-related VACs were released in DR16+ (a mini data release that happened in 2020 July). In addition, a version of the "Visual Morphology from DECaLS Images" VAC, which is updated for DR17, was also released in DR16+. We document those first.

• 1.  
  NASA Sloan Atlas Images and Image Analysis: This VAC contains the underlying image and image analysis for the NSA. The methods used are described in Blanton et al. (2011) and Wake et al. (2017). Briefly, for a set of nearby galaxies of known redshift (z < 0.15) within the SDSS imaging area, we have created and analyzed GALEX (Morrissey et al. 2007) and SDSS images. This analysis forms the basis for the MaNGA targeting, and resulted in the v1_0_1 NSA catalog released originally with DR14. We are now releasing the images that were analyzed to create those parameters. The data set includes the original catalogs from which the NSA sample was drawn, the mosaic images and inverse variance images that were analyzed, the deblending results for each object, the curve-of-growth and aperture corrections for each object, and other intermediate outputs. We expect that this data set may be useful for reanalysis of the GALEX or SDSS imaging. The full data set is large (15 TB), and therefore any users interested in using a large fraction of it should transfer the data through Globus (see Section 3 for details on how to use Globus ¹⁷⁸ ).
• 2.  
  MaNGA SWIFT VAC: The Swift+MaNGA VAC comprises 150 galaxies with both SDSS-IV/MaNGA IFU spectroscopy and archival Swift Ultraviolet Optical Telescope (UVOT) near-ultraviolet (NUV) images, and is presented in Molina et al. (2020). The similar angular resolution (∼3'') between the Swift/UVOT three NUV filters and the MaNGA IFU maps allows for spatially resolved comparisons of optical and NUV star formation indicators, which is crucial for constraining attenuation and star formation quenching in the local universe. The UVOT NUV images, SDSS optical images, and MaNGA emission-line and spectral index maps have been spatially matched and reprojected so that all of the data match the pixel sampling, resolution, and coordinate system of the UVOT uvw2 image for each galaxy. The spectral index maps utilize the definition given in Burstein et al. (1984), which allows users to more easily compute spectral indices when binning the maps. Spatial covariance is properly accounted for in the propagated uncertainties. In addition to the spatially matched maps, Molina et al. (2020) also provide a catalog with PSF-matched aperture photometry for the SDSS optical and Swift/UVOT NUV bands.

A variety of galaxy morphology catalogs are provided as VACs, with analysis done in a variety of ways, using a variety of images. We provide a short summary of each here—for more details, please see the appropriate paper.

• 1.  
  Galaxy Zoo: 3D (Masters et al. 2021): This provides crowdsourced spaxel masks locating galaxy centers, foreground stars, bars, and spirals in the SDSS images of MaNGA target galaxies. Available for use within Marvin, these masks can be used to pick out spectra, or map quantities likely associated with the different structures (see Fraser-McKelvie et al. 2019, 2020; Peterken et al. 2019a, 2019b; Greener et al. 2020; Krishnarao et al. 2020, for example use cases).
• 2.  
  Galaxy Zoo Morphologies from SDSS, DECaLS, and UKIDSS : The Galaxy Zoo method, which involves combining classifications from a large number of classifiers collected via an online interface, has been applied to a variety of images, including the original SDSS images (Willett et al. 2013; Hart et al. 2016), the UKIDSS (Lawrence et al. 2007; Galloway 2018), and most recently the Dark Energy Camera Legacy Survey (DECaLS; Dey et al. 2019; Walmsley et al. 2021). This latter analysis combines machine learning (ML) methods with crowdsourcing in an active loop (for details see Walmsley et al. 2022). We collect together all these crowdsourced morphologies for as many MaNGA galaxies as possible in this VAC.
• 3.  
  Visual Morphology from DECaLS Images : This VAC contains a direct visual morphological classification, based on the inspection of image mosaics generated from a combination of SDSS and DECaLs (Dey et al. 2019) images, for the MaNGA galaxies. The DR16+ version contains the classification for the first half of MaNGA galaxies (4600, MaNGA DR15) while the DR17 version contains the classification for the full MaNGA DR17 with unique MaNGAID. Through digital image postprocessing, we exploit the advantages of this combination of images to identify inner structures, as well as external low-surface-brightness features for a homogeneous classification, following an empirical implementation of the methods in Hernández-Toledo et al. (2010) and Cheng et al. (2011). The visual morphological classification is carried out by two classifiers inspecting three-panel image mosaics, containing a gray logarithmic-scaled r-band image, a filter-enhanced r-band image, and the corresponding r, g, and b color composite image from SDSS; and a similar mosaic using DECaLS images incorporating the residual image after subtraction of a best surface brightness model from the DESI legacypipeline. ¹⁷⁹ The catalog contains the T-Type morphology, a variety of visual morphological attributes (bars, bar families, tidal debris, etc.), and our estimate of the non-parametric structural, concentration, asymmetry, and clumpiness (see CAS; Conselice 2003) parameters from the DECaLS images. For more detail, see Vázquez-Mata et al. (2021). An updated version including morphologies for all DR17 MaNGA galaxies is being prepared; see J. A. Vázquez-Mata et al. (2022, in preparation).
• 4.  
  MaNGA PyMorph DR17 Photometric Catalog (MPP-VAC; see Fischer et al. 2019; Domínguez Sánchez et al. 2021 for details): This provides photometric parameters obtained from Sèrsic and Sèrsic+Exponential fits to the 2D surface brightness profiles of the final MaNGA DR17 galaxy sample (e.g., total fluxes, half light radii, bulge-disk fractions, ellipticities, position angles, etc.). It extends the MaNGA PyMorph DR15 photometric VAC to now include all MaNGA galaxies in DR17.
• 5.  
  MaNGA Morphology Deep Learning DR17 Catalog (MDLM-VAC; see Domínguez Sánchez et al. 2021 for details): This provides morphological classifications for the final MaNGA DR17 galaxy sample using convolutional neural networks. The catalog provides a T-Type value (trained in regression mode) plus four binary classifications: P_LTG (separates early type galaxies, or ETGs, from late types, or LTGs) , P_S0 (separates ellipticals from S0), P_edge‐on (identifies edge-on galaxies), and P_bar (identifies barred galaxies). It extends the "MaNGA Deep Learning Morphology DR15 VAC" (Fischer et al. 2019) to now include galaxies that were added to make the final DR17. There are some differences with respect to the previous version—namely, the low-end of the T-Types are better recovered in this new version. In addition, the P_LTG classification separates ETGs from LTGs in a cleaner way, especially at the intermediate types (−1 < T-Type < 2), where the T-Type values show a large scatter. Moreover, the value provided in the catalog is the average of 10 models trained with k-folding for each classification task (15 for the T-Type classification). The standard deviation, which can be used as a proxy for the uncertainty in the classification, is also reported.

There are a variety of stellar population modeling-based VACs released.

• 1.  
  Principle Component Analysis (PCA) VAC (DR17): This VAC includes measurements of resolved and integrated galaxy stellar masses, obtained using a low-dimensional, PCA-derived fit to the stellar continuum and subsequent matches to simulated star formation histories. The general methodology for obtaining the principal component basis set, the stellar continuum fitting routine, and the process of inferring stellar population properties such as mass-to-light ratio are discussed in Pace et al. (2019a). The aggregation of pixel-based mass estimates and the adopted aperture-correction procedure are described in Pace et al. (2019b). This procedure yields estimates of galaxy-wide, integrated stellar masses also provided as part of this VAC. Key VAC characteristics remain unchanged in comparison to DR16 (Ahumada et al. 2020), where a holistic description of the VAC can be found. The principal enhancement in this release is in the sample size: the number of galaxies has been expanded to include all MaNGA galaxies to which the analysis could be readily applied, a total of 10,223 unique plate-ifu designations (this number differs from unique galaxy counts, as some MaNGA galaxies were observed multiple times, so they have multiple plate-ifus, each of which are analyzed separately in this VAC).
• 2.  
  Firefly Stellar Populations: This VAC provides measurements of spatially resolved stellar population properties of MaNGA galaxies employing the Firefly ¹⁸⁰ (Wilkinson et al. 2017) full-spectral fitting code. For DR17, Firefly v1.0.1 was run, over all 10,735 data cubes that had been processed by both the DRP and the DAP. ¹⁸¹ The major difference to the DR15 VAC is that we now provide the catalog in two versions. The first employs the stellar population models of Maraston & Strömbäck (2011) based on the MILES stellar library (Sánchez-Blázquez et al. 2006). The second version uses new MaStar models described in Maraston et al. (2020). Both model libraries assume a Kroupa (2001) IMF. Compared to the Firefly VAC in DR15, the radius (stored in HDU4 in the file) is now given in elliptical coordinates, and the azimuth is added. Masses (HDU11 and HDU12) are given per spaxel and per Voronoi cell. We do not provide absorption index measurements anymore. Each version of the VAC is offered as a single FITS file (∼6 GB) comprising the whole catalog of global and spatially resolved parameters and, additionally, as a small version (∼3 MB) that contains only global galaxy stellar population parameters. A detailed description can be found in Neumann et al. (2022) and Goddard et al. (2017).
• 3.  
  Pipe3D: This VAC contains the Pipe3D (Sánchez et al. 2016) analysis of the full MaNGA data set comprising the main properties of the stellar populations and emission lines for more than 10,000 galaxies, both spatial resolved and integrated across the entire FOV of the IFUs. The content of the released distribution was originally described in Sánchez et al. (2018), and updated in S. F. Sánchez et al. (2022, in preparation). The new releases include considerable modifications from the previous ones, the most important ones being (i) the use of an updated version of the code fully transcribed to Python (E. Lacerda et al. 2022, in preparation), (ii) the use of a new stellar population library based on the MaStar stellar library (A. Mejia-Narvaez et al. 2022, in preparation), and (iii) an update on the list of analyzed emission lines.

H i-MaNGA is a H i 21 cm line follow-up program, to provide estimates of total atomic hydrogen content for galaxies in the MaNGA survey (Masters et al. 2019). It makes use of both previously published H i data (primarily from the ALFALFA survey; Haynes et al. 2018) and new observations using the Robert C. Bryd Green Bank Telescope (GBT; to date under observing codes GBT16A_095, GBT17A_012, GBT19A_127, GBT20B_033 and GBT21B_130). This VAC comprises the third data release (DR3) of H i 21 cm detections or upper limits for 6358 galaxies in the MaNGA sample. In some cases, both GBT and ALFALFA data exist for a single galaxy, and we provide both observations separately, so the total number of rows is 6632, with 3358 coming from our GBT observations. The observation and reduction strategies are documented in Masters et al. (2019), Stark et al. (2021). As part of this program, a 20% offset between actual and estimated L-band calibration at GBT was noticed (see Goddy et al. 2020). Stark et al. (2021) provide guidance on dealing with confusion, and including upper limits into statistical analysis. Observations are ongoing (under proposal code GBT21B_130), with the program on track to observe, or homogenize H i data for all MaNGA galaxies at z < 0.05, with no preselection on color or morphology. This is expected to result in a final H i-MaNGA sample size of around 7000 MaNGA galaxies, with over 6800 already having at least some data in hand.

The MaNGA AGN Catalog presents AGN in the DR15 sample of MaNGA that are identified via mid-infrared WISE colors, Swift/BAT ultrahard X-ray detections, NRAO-VLA Sky Survey, and Faint Images of the Radio Sky at Twenty-Centimeters (FIRST) radio observations, and broad emission lines in SDSS spectra. The catalog further divides the radio AGN into quasar-mode and radio-mode subpopulations, and provides estimates of the AGN bolometric luminosities. Full details of the AGN selection and luminosity measurements are described in Comerford et al. (2020). It is intended that this will be updated to include all the MaNGA galaxies in the future.

The Galaxy Environment for MaNGA (GEMA) VAC (M. Argudo-Fernádez et al. 2022, in preparation) provides a variety of different measures of environment for galaxies in MaNGA. The combination of mass-dependent and mass-independent parameters provided in the catalog can be used to explore the effects of the local and large-scale environments on the spatial distribution of star formation enhancement/quenching, in the interaction of AGN with galaxies, or in the connection with kinematics or galaxy morphology, for instance, in a homogenous way allowing comparisons between different studies. In DR17, we present the final and updated version of GEMA-VAC for the final MaNGA sample. The quantifications of the environments are based on the methods described in Argudo-Fernández et al. (2015) to estimate tidal strengths and projected number densities, as well as that in Etherington & Thomas (2015) to estimate overdensity-corrected local densities (MaNGA galaxies in the SDSS-DR15 only), and Wang et al. (2016) for an estimation of the cosmic web environment. To better explore the environment of galaxies located in dense local environments, for instance, galaxies in compact groups or with strong interactions (close paris/mergers), but not necessarily a high-density environment at larger scale, we also provide these same quantifications considering MaNGA galaxies in groups, according to an updated version of the catalog of groups compiled by Yang et al. (2007); and MaNGA galaxies in close pairs, according to the sample used in Pan et al. (2019).

We present a catalog of precise spectroscopic redshifts (M. Talbot et al. 2022, in preparation) for the RSS and spaxels in MaNGA, updating the previous version of this catalog (Talbot et al. 2018) to include the completed sample of MaNGA galaxies. These spectroscopic redshifts are computed using the spec1d—zfind code from the publicly available BOSS pipeline (Bolton et al. 2012), in which the NSA catalog provides the initial redshift. Once spectroscopic redshifts were determined for the high-S/N region within the galaxy half-light radius, a second pass attempted to determine the remaining redshifts in the low-S/N spectra using the mean spectroscopic redshift as a prior. The spectroscopic redshifts and a foreground model are presented for each spectrum with sufficient S/N to model, in this VAC.

We present 6 likely, 12 probable, and 74 possible candidates as strong galaxy–galaxy scale gravitational lenses found within the completed MaNGA survey. The lens candidates are found by the Spectroscopic Identification of Lensing Objects (SILO) program (Talbot et al. 2018, 2021; M. Talbot et al. 2022, in preparation), which was adapted from the BOSS Emission-Line Lens Survey (BELLS; Brownstein et al. 2012) spectroscopic detection method to find background emission lines within coadded foreground-subtracted row-stacked-spectra of the MaNGA IFU, in which the coadded residuals are stacked across exposures from the same fiber at the same dither position. Visual inspection of any background emission line detected was performed, including the position of detections in proximity to an estimated Einstein radius. Narrowband images were constructed from the coadded residuals to search for any lensing features.

DR16 included full reductions of the completed set of the observed eBOSS spectra. An additional series of redshift estimates for the eBOSS galaxy samples was produced by an algorithm known as redrock. ¹⁸⁴ The galaxy redshifts derived from redrock are described in Section 4 of Ross et al. (2020). An additional series of classifications and redshift estimates was also performed for the BOSS and eBOSS quasar samples. The origin of quasar redshift estimates is described in Lyke et al. (2020), while a summary of how those redshifts were used in the clustering measurements is also found in Section 4 of Ross et al. (2020). From these updated redshift estimates, LSS VACs are created, which, together map the three-dimensional structure of the universe using galaxies and quasars over redshifts 0.6 < z < 2.2. These maps are carefully constructed with corrections for observational systematic errors and random positions that sample the survey selection function to allow unbiased cosmological inference. Three distinct samples were observed by SDSS-IV and used to produce LSS catalogs: LRG (Prakash et al. 2016), ELG (Raichoor et al. 2017), and quasars (Myers et al. 2015). A VAC for each of these samples was released in 2020 July. The LSS catalogs for the LRG and quasar samples are described in Ross et al. (2020) while the LSS catalog for the ELG sample is described in Raichoor et al. (2021).

We present 1000 realizations of multi-tracer EZmock catalogs, with redshift evolution and observational systematics, for each sample of the DR16 LSS data. These mock catalogs are generated using the EZmock method (Chuang et al. 2015), and applied the survey footprints and redshift distributions extracted from the corresponding data. They accurately reproduce the two- and three-point clustering statistics of the DR16 data, including cross-correlations between different tracers, down to the scale of a few h⁻¹ Mpc, and provide reliable estimates of covariance matrices and analyses on the robustness of the cosmological results. Details on the construction and clustering properties of the EZmock catalogs are presented in Zhao et al. (2021).

Beginning with SDSS-I, SDSS has maintained a tradition of releasing a visually inspected quasar (or quasi-stellar object; QSO) catalog alongside major data releases. The new SDSS-DR16Q catalog (DR16Q; Lyke et al. 2020) represents the most recent, and largest, catalog of known unique quasars within SDSS. To ensure completeness, the quasars from previous catalog releases (DR7Q; Schneider et al. 2010; DR12Q; Pâris et al. 2017) have been combined with observations from eBOSS in SDSS-IV. The catalog contains data for more than 750,000 unique quasars, including redshifts from visual inspections, PCA, and the SDSS automated pipeline. Additionally, the catalog is the first from SDSS to contain both the Hewett & Wild (2010) DR6 redshift estimates and the Shen et al. (2011) DR7 redshift estimates that are based on the Hewett & Wild (2010) algorithm. Where applicable, the catalog also contains information about broad absorption line troughs, damped Lyα absorbers, and emission-line redshifts (via PCA). As in previous releases, DR16Q also contains properties for each quasar from GALEX (Martin et al. 2005), UKIDSS (Lawrence et al. 2007), WISE (Wright et al. 2010), FIRST (Becker et al. 1995), 2MASS (Skrutskie et al. 2006), ROSAT/2RXS (Boller et al. 2016), XMM-Newton (Rosen et al. 2016), and Gaia (Gaia Collaboration et al. 2018), when available. To facilitate analyses of pipeline accuracy and automated classification, a superset was also released. This sample contained ∼1.4 million unique observations for the objects targeted as quasars from SDSS-I/II/III/IV.

This VAC contains the estimated fluctuations of the transmitted flux fraction in the pixels across the Lyα and Lyβ spectra region of DR16Q quasars. In total, 211,375 line of sights contribute to the Lyα spectral regions and 70,626 to the Lyβ one. This VAC contains everything needed to compute the three-dimensional auto-correlation of Lyα absorption in two different spectral regions as in du Mas des Bourboux et al. (2020). When combined with the DR16Q quasar catalog, this VAC also provides the information to compute the three-dimensional quasar × Lyα cross-correlation. These two measurements are used to measure the location of the BAO as reported in du Mas des Bourboux et al. (2020).

A VAC of 838 likely, 448 probable, and 265 possible candidate strong galaxy gravitational lens systems was released along with the batch of cosmology results and VACs in 2020 July (aka DR16+). These systems were discovered by the presence of higher-redshift-background emission lines in DR16 eBOSS galaxy spectra. The methodology, including quantitative explanations of the "likely," "probably," and "possible" categorization is described in full in Talbot et al. (2021); also see Talbot et al. (2018). This SILO program extends the method of the BELLS (Brownstein et al. 2012) and Sloan Lens ACS (Bolton et al. 2006) survey to a higher redshift, and has recently been applied to the spectroscopic discovery of strongly lensed galaxies in MaNGA (see SILO; Talbot et al. 2018; also see our Section 5.5.8). Although these candidates have not been studied through a dedicated imaging program, an analysis of the existing imaging from the SDSS Legacy Survey and the DESI Legacy Surveys (Dey et al. 2019) provides additional information on these systems. This catalog includes the results of a manual inspection process, including grades and comments for each candidate, consideration of sky contamination, low-S/N emission lines, improper calibration, weak target emission lines, systematic errors, Gaussian modeling, and potential lensing features visually identified within the available imaging.

The eBOSS ELG-LAE Strong Lens Catalog contains roughly 150 candidate strong lens systems selected from the ELG sample of the eBOSS survey released by DR17 using the method presented in Shu et al. (2016). By construction, the lensing galaxies in this catalog are ELGs up to z ≃ 1.1, and the source galaxies are Lyα emitters at z > 2. A full description of the catalog will be presented in A. Filipp et al. (2022, in preparation). This catalog is complementary to the eBOSS Strong Gravitational Lens Detection Catalog constructed by Talbot et al. (2021, Section 7.2.1 above), in which the source galaxies are all [O ii] emitters at z < 1.7.

The Monte Carlo Physarum Machine (MCPM) cosmic web reconstruction algorithm (Burchett et al. 2020; Elek et al. 2021) was employed to characterize the matter density field from galaxies spectroscopically observed in SDSS, including those previously included in the NSA (Blanton et al. 2011) and LOWZ and eBOSS LRG catalogs. The MCPM framework, which is particularly sensitive to the filamentary structure of the cosmic web, requires as inputs galaxy coordinates, redshifts, and halo masses, and we leverage stellar mass measurements included in VACs from previous data releases, the NSA itself, and the eBOSS Firefly VAC (Comparat et al. 2017). Halo masses were then derived by the halo abundance matching relations from Moster et al. (2013). Using the Polyphorm software (Elek et al. 2020), in which MCPM is implemented, reconstructions were produced in various redshift ranges corresponding to the varying depth/completeness of the galaxy samples. MCPM then provides a proxy metric (known as the trace) for the environmental density at each point in the fitted volume. These arbitrary density values are then calibrated to the cosmological matter density relative to the mean matter density by performing MCPM fits to the Bolshoi–Planck (Klypin et al. 2016) dark-matter-only cosmological simulation's halo catalog (Behroozi et al. 2013) and producing a mapping between the MCPM trace and simulation 3D matter density field. The data products are as follows: (1) galaxy catalogs with local environmental density values evaluated at their locations from the full 3D density field and (2) the 3D density field itself, which users may in turn query for density values at locations away from known galaxy positions.

The identification of SPIDERS clusters and their membership, as well as the detailed target selection process, are described fully by A. Merloni et al. (2022, in preparation) and J. Ider Chitham et al. (2022, in preparation). Here we document the resulting targeting bits that describe the contents of the SPIDERS cluster target catalog:

• 1.  
  EFEDS_EROSITA_CLUS: These are galaxies associated with extended X-ray emission. A combination of all member galaxies selected using redMaPPer in scanning-mode (Rykoff et al. 2014) as well as additional BCGs identified using the multicomponent matched filter cluster confirmation tool (Klein et al. 2019). BCGs are allocated with the highest-target priority (0), while all other members are given a lower priority (33-130). The selection method on eFEDS extended sources is described fully in Liu et al. (2021a), Klein et al. (2021). The redMaPPer-based selection uses positional X-ray information to estimate optical properties such as photometric redshifts, centering information, and member selection in a method analogous to that of Clerc et al. (2016), as used in the SPIDERS DR16 data release. The method itself is described by Ider Chitham et al. (2020), J. Ider Chitham (2022, in preparation). The photometric selections of BCGs and cluster members are based on the eighth data release of the DESI Legacy Imaging Surveys (using g − r and r − z colors) as well as HSC photometric (grizy) data (Aihara et al. 2019; Dey et al. 2019). Cluster BCG photometric redshifts are within 0.0 < z < 1.3, and a total of 85 were observed.
• 2.  
  EFEDS_SDSS_REDMAPPER: Publicly available optically selected cluster member catalog based on SDSS DR8 and CODEX data (Rykoff et al. 2014; Finoguenov et al. 2020). Cluster photometric redshifts are within 0.0 < z < 0.6, and a total of 1031 potential cluster galaxies were observed.
• 3.  
  EFEDS_HSC_REDMAPPER: Optically selected cluster member catalog (Rykoff et al. 2014) based on HSC DR2 data (Aihara et al. 2019). Cluster photometric redshifts are within 0.3 < z < 1.1, and a total of 13 cluster galaxies were observed.

The method used to associate optical/IR counterparts to the eFEDS X-ray point-sources is presented by Salvato et al. (2021). In short, the optical counterparts to the X-ray sources were identified using a Bayesian association algorithm NWAY (Salvato et al. 2018), which accounts for both position and photometric properties of the counterpart. The association algorithm was informed by a large training sample of XMM-Newton and Chandra X-ray serendipitous sources with secure optical identifications. Supplementary optical/IR counterparts were also provided via a Likelihood Ratio approach (Sutherland & Saunders 1992; Ruiz et al. 2018). Note that, at the time of selection of targets for the SDSS-IV special plates, only very early versions of the X-ray catalog and counterpart selection algorithms were available—so the pool of targets is different from the counterparts presented by Salvato et al. (2021). The main eFEDS pointlike target class is EFEDS_EROSITA_AGN with 2149 observed targets. Despite the name, we expect this target set to also include a number of stars and (inactive) galaxies.

To ensure that all the fibers available in the designed plates could be allocated to interesting scientific targets, beyond the eROSITA-selected sources described above, several additional target classes were defined before the arrival of the eROSITA data. The largest of these were special classes of AGN drawn from optical targets selected by a combination of the HSC Subaru Strategic Program imaging and other multiwavelength catalogs. Note that, due to the high sky density seen in the available eROSITA/eFEDS source catalog, very few fibers were left for filler targets, and so most of the defined target classes were allocated zero fibers (unless the same astrophysical object was also selected as an eROSITA target). However, since these target selection classes are defined in the DR17 data products, for completeness of the documentation, we give a brief description here of each target class.

• 1.  
  EFEDS_HSC_HIZQSO: Optical/NIR High-redshift QSO candidates selected from HSC photometric data with a drop-out technique targeting z > 4
• 2.  
  EFEDS_HSC_WERGS: FIRST-detected bright radio galaxies selected for their extreme red Optical-NIR colors
• 3.  
  EFEDS_HSC_REDAGN: WISE-detected obscured AGN candidates selected on the basis of their extremely red Optical-NIR colors
• 4.  
  EFEDS_COSMOQSO: UV excess/variability selected QSO—following prototype selection criteria of 4MOST Cosmology Survey
• 5.  
  EFEDS_WISE_AGN: AGN candidate selected via location in WISE color–magnitude space, following criteria of Assef et al. (2018)—with optical counterparts selected from DESI Legacy Imaging Survey DR8
• 6.  
  EFEDS_WISE_VARAGN: AGN candidate selected via variability signature in WISE data—with optical counterparts selected from DESI Legacy Imaging Survey DR8
• 7.  
  EFEDS_XMMATLAS: Optical counterparts to X-ray sources from the 3XMM/dr8 catalog (Rosen et al. 2016) and from the XMM-ATLAS survey field (Ranalli et al. 2015)
• 8.  
  EFEDS_CSC2: X-ray sources from the Chandra-CSC2.0 catalog (Evans et al. 2010)—with optical counterparts selected from DESI Legacy Imaging Survey (DR8)
• 9.  
  EFEDS_SDSSWISE_QSO: QSO candidates from the Clarke et al. (2020) ML classification of SDSS-WISE photometric sources
• 10.  
  EFEDS_GAIAWISE_QSO: QSO candidate from the Shu et al. (2019) Random Forest classification of Gaia-unWISE sources
• 11.  
  EFEDS_KIDS_QSO: QSO candidate from the Nakoneczny et al. (2019) classification of KiDS+VIKING DR3 sources
• 12.  
  EFEDS_GAIA_WD: Gaia-selected WD binary candidate in the region between the zero age main sequence and the WD sequence.
• 13.  
  EFEDS_GAIAGALEX_WD: Gaia-GALEX UV excess source—expected to be compact binary harboring a WD
• 14.  
  EFEDS_VAR_WD: WD candidate selected on the basis of its optical variability

Further details of the eFEDS target selection process will be presented in A. Merloni et al. (2022, in preparation).

The targets described above for the eFEDS field are assigned to 19 different target classes. Table 5 describes each of the target classes designed for the 12 plates covering eFEDS. The maskbits used range from 19 to 47 and can be found in the EBOSS_TARGET1 target bitmask. For clarity, only the bits for target classes that resulted in observations are reported.

Observations were carried out in 2020 March, but due to a combination of poor weather and COVID-19 shut down (on MJD 58,932), the program could unfortunately not be completed. The initial success criteria for a plate to be considered complete were SN2_G1,2 > 20, SN2_I1,2 > 40, where the SN2 are estimates of the squared S/N, at fiducial 2'' fiber magnitudes of g = 21.2 and i = 20.2, averaged over the targets in each of the four BOSS spectrograph arms (G1, I1, G2, I2). These SN2 thresholds are approximately double those adopted for regular eBOSS observations (Dawson et al. 2016). This guarantees highly complete redshift measurements for faint AGN, and allows the measurement of the properties of the AGN and of the host galaxy.

Table 6 details the plate numbers and the meta data related to their observations. Only seven out of 12 plates were actually observed in the period MJD 58,928–58,932; of these seven, only one plate reached nominal depth, three received significant exposure beyond the typical eBOSS limit, and three were only partially exposed. In Table 6, these are marked as good ("G"), fair ("F"), and bad ("B"), respectively.

The redshifts from eFEDS special plates are part of the catalog that describes the multiwavelength properties of the eROSITA point sources (Salvato et al. 2021), currently available at https://erosita.mpe.mpg.de/edr/eROSITAObservations/Catalogues/ and expected to be available in Vizier in the future.

The spectroscopic observations of the eROSITA sources in the eFEDS PV field, reduced at the end of the SDSS-IV program because of the 2020 March closure at APO, has been resumed within the early (plate-mode operations) phase of SDSS-V (see Section 8). A much more complete description of the eFEDS program (SDSS-IV and SDSS-V) will be provided in A. Merloni et al. (2022, in preparation). Among others, the more homogeneous coverage of the combined program will enable the X-ray AGN clustering studies and galaxy–galaxy lensing measurements (J. Comparat et al. 2021, in preparation) on an unprecedentedly large sample.

Eventually, with the commissioning of the robotic fiber positioners on both northern and southern SDSS-V sites, SPIDERS will deliver on its original promises of massive, systematic spectroscopic observations of the sources detected in the eROSITA all-sky survey as part of the SDSS-V BHM program (Kollmeier et al. 2017). The experience and the scientific results obtained by the SPIDERS team within SDSS-IV, which we have briefly described above, represent a major milestone for the planning, execution, and exploitation of the BHM survey program.